Neck-wearable communication device with microphone array

Abstract

A wearable electronic device includes a neck-wearable housing, generally U-shaped, with an electrical connector; two in-ear earphones; two cords, with one end connected to one of the earphones and the other end connected to the connector. The two cords are mechanically connected to the housing. Points of connection of the cords to the housing are close to each other and form a dorsal node, and the two cords are mechanically connected to each other in their portions between the in-ear earphones and the dorsal node to form a suboccipital node. A microphone is placed in the housing on a front (chest) side of the user. A microphone is placed in the dorsal and/or in the suboccipital node. The microphones are used for determination of correlated and non-correlated components of audio signals. The correlated components are treated as a noise signal and the non-correlated components are a target signal.

Claims

1. A headset for a mobile electronic device, comprising: a neck-wearable housing having a generally U-shape with an electrical connector attached thereto; two in-ear earphones; two cords, each connected at one end to a corresponding in-ear earphone and connected at its other end to the electrical connector; and at least one noise reduction microphone array disposed on the mobile electronic device, wherein the two cords are mechanically connected to the neck-wearable housing, and points of connection of the cords to the neck-wearable housing are in close proximity to each other and form a dorsal cord connection node, and wherein the two cords are also mechanically connected to each other in sections between the in-ear earphones and the dorsal cord connection node to form a suboccipital cord connection node at the connection point.

2. The headset of claim 1, wherein the dorsal cord connection node and the suboccipital cord connection node are located on a dorsal surface of a neck, and cords in sections between the in-ear earphones and the suboccipital node are located over an auricle.

3. The headset of claim 1, wherein the suboccipital cord connection node is a clip adapted to move along the two cords for adjusting a length of the two cords.

4. The headset of claim 1, wherein the suboccipital node comprises an electrical connector for disconnecting the cords.

5. The headset of claim 1, wherein at least one of the two cords is a helical spring in a section between the suboccipital and dorsal cord connection nodes.

6. The headset of claim 1, further comprising an electronic unit mechanically and electrically coupled to the electrical connector.

7. The headset of claim 6, further comprising buttons disposed on the neck-wearable housing for control of the electronic unit.

8. The headset of claim 6, further comprising at least one power supply for the electronic unit, the power supply disposed on the neck-wearable housing.

9. The headset of claim 1, wherein the neck-wearable housing is flexible in at least one location.

10. A wearable telecommunication device, comprising: a neck-wearable housing with an electronic unit attached thereto; two in-ear earphones; two cords, one of which connects to one of the in-ear earphone to the electronic unit, and the other cord connects the other in-ear earphone to the electronic unit; and a microphone array for picking up and processing a user's voice, the microphone array comprising a front microphone, a rear microphone and a processor; wherein the two cords are mechanically connected to the neck-wearable housing, and points of connection of the two cords to the neck-wearable housing are close to each other and form a dorsal cord connection node, and are further mechanically connected to each other in sections between the in-ear earphones and the dorsal cord connection node to form a suboccipital cord connection node; wherein the suboccipital cord connection node, the dorsal cord connection node, and an area of the housing close to the dorsal cord connection node form a rear portion of the wearable telecommunication device; wherein the rear microphone is fixed on the rear portion; and wherein a front-facing portion of the neck-wearable housing is in contact with an upper chest when worn by the user.

11. The device of claim 10, wherein the rear microphone is fixed on the suboccipital cord connection node.

12. The device of claim 10, wherein the rear microphone is fixed on the neck-wearable housing close to the dorsal cord connection node.

13. The device of claim 10, wherein the rear microphone is between the suboccipital cord connection node and the dorsal cord connection node.

14. The device of claim 10, further comprising a spring between the suboccipital cord connection node and the dorsal cord connection node.

15. The device of claim 14, wherein the rear microphone is fixed on the spring.

16. The device of claim 10, wherein the rear microphone detects surrounding noise, and wherein a correlated portion of the signal from the rear microphone and the signal from the front microphone represents noise, while an uncorrelated portion of the signal represents a useful data.

17. The device of claim 16, wherein signals from the rear microphone and the front microphone are ignored when their correlation is above a pre-defined threshold.

18. The device of claim 10, wherein the front microphone is fixed on the front-facing portion of the neck-wearable housing.

19. The device of claim 10, wherein the microphone array includes at least two front microphones, wherein the two front microphones are fixed on the front-facing portion of neck-wearable housing at a substantially the same height when worn by the user and one of the at least two front microphones is close to or below a right clavicle of the user, and the other of the at least two front microphones is close to or below a left clavicle of the user.

20. The device of claim 10, further comprising at least one gradient microphone array comprising at least two front microphones fixed on the front-facing portion of the neck-wearable housing at different heights when worn by the user, wherein the gradient microphone array is used to determine a directional diagram of received sound waves.

21. The device of claim 10, further comprising at least one phased microphone array comprising at least two front microphones fixed on the front-facing portion of the neck-wearable housing at different heights when worn by the user, wherein the phased microphone array is used to determine a directional diagram of received sound waves.

22. The device of claim 10, further comprising an electronic accessory in a form of a wrist watch, which is wirelessly connected to the electronic unit, wherein the electronic accessory includes the front microphone.

23. The device of claim 10, further comprising an electronic accessory in a form of a finger ring, which is wirelessly connected to the electronic unit, wherein the electronic accessory includes the front microphone.

24. The device of claim 10, further comprising an electronic accessory in a form of eyeglasses, which is connected to the electronic unit, wherein the electronic accessory includes the front microphone.

25. The device of claim 10, wherein the neck-wearable housing is generally U-shaped.

26. The device of claim 10, wherein the neck-wearable housing is generally O-shaped.

27. The device of claim 10, wherein the neck-wearable housing is flexible in at least one location.

28. A wearable telecommunication device, comprising: a neck-wearable housing configured to be mounted on a human body and in contact with back, left, right sides of the neck and upper chest and having at least one electronic unit attached thereto; two in-ear earphones, two cords, one of which connects to one of the in-ear earphone to the electronic unit, and the other cord connects the other in-ear earphone to the electronic unit, a microphone array for picking up and processing a user's voice, comprising a front microphone, a rear microphone and processor; wherein the two cords are mechanically connected to the neck-wearable housing and points of connection of the two cords to the neck-wearable housing are close to each other and form a dorsal cord connection node, and are further mechanically connected to each other in sections between the in-ear earphones and the dorsal cord connection node to form a suboccipital cord connection node at the connection point; wherein the rear microphone is on a portion of the neck-wearable housing configured to be in contact with a back of the neck when worn by the user; and wherein a correlated portion of the signal from the rear microphone and the signal from the front microphones represents noise, while an uncorrelated portion of the signal represents useful data.

29. The device of claim 28, wherein the rear microphone is fixed on any of (i) the suboccipital cord connection node, (ii) the neck-wearable housing close to the dorsal cord connection node, and (iii) between the suboccipital cord connection node and the dorsal cord connection node.

30. The device of claim 28, further comprising a spring between the suboccipital cord connection node and the dorsal cord connection node, wherein the rear microphone is on the spring.

31. The device of claim 28, wherein the front microphone is fixed on a portion of the neck-wearable housing that is in contact with the user's chest when worn by the user.

32. The device of claim 28, wherein signals from the rear microphone and the front microphone are ignored when their correlation is above a pre-defined threshold.

33. The device of claim 28, wherein the device forms an output signal e.sub.n as:
e.sub.n=d.sub.n−y.sub.n, where y.sub.n is a correlated signal representing filtered noise calculated as:
y.sub.n=w.sub.n.sup.Tx.sub.n, where x.sub.n is a combined signal from the rear microphone, d.sub.n is a combined signal from the front microphones, w.sub.n are adaptive filter coefficients defined as:
w.sub.n+1=w.sub.n+μe.sub.nx.sub.n, where w.sub.n+1 is a set of coefficients at a current moment of time n+1, w.sub.n is a set of coefficients at a previous moment of time, n is defined by a clock rate of the incoming data stream, μ is a positive value defining stability and convergence rate.

34. The device of claim 28, wherein the device forms an output signal e.sub.n based on a Filtered-X Least-Mean-Square (FXLMS) algorithm:
e.sub.n=d.sub.n−P(z)y.sub.n, where d.sub.n is a combined signal from the front microphones, P(z) is a transfer function, and y.sub.n is a correlation signal defined by:
y.sub.n=w.sub.n.sup.Tx.sub.n, where x.sub.n is a combined signal from the rear microphone, w.sub.n are adaptive filter coefficients defined as:
w.sub.n+1=w.sub.n+μe.sub.nr.sub.n, where w.sub.n+1 is a set of coefficients at a current moment of time n+1, w.sub.n is a set of coefficients at a previous moment of time n, n is defined by a clock rate of the incoming data stream, μ is a positive value defining stability and convergence rate, r.sub.n is a vector formed by current and previous values of a filtered base signal r.sub.n according to:
r.sub.n={circumflex over (P)}(z)x.sub.n, where {circumflex over (P)}(z) is based on a Least Mean Squares Algorithm that corresponds to a finite impulse response (FIR) filter.

35. The device of claim 28, wherein the microphone array includes two front microphones, wherein the two front microphones are fixed on a portion of the neck-wearable housing that is in contact with the user's chest and at a substantially the same height when worn by the user and one of the two front microphones is close to or below a right clavicle of the user, and the other of the two front microphones is close to or below a left clavicle of the user.

36. The device of claim 28, further comprising at least one gradient microphone array comprising at least two front microphones fixed on the front-facing portion of the neck-wearable housing at different heights when worn by the user, wherein the gradient microphone array is used to determine a directional diagram of received sound waves.

37. The device of claim 28, further comprising at least one phased microphone array comprising at least two front microphones fixed on the front-facing portion of the neck-wearable housing at different heights when worn by the user, wherein the phased microphone array is used to determine a directional diagram of received sound waves.

38. The device of claim 28, wherein the rear microphone is used as a detector of surrounding noise wherein a correlated portion of the signal from the rear microphone and the signal from the front microphone represents noise, while an uncorrelated portion of the signal represents a useful data.

39. The device of claim 28, wherein signals from the rear microphone and the front microphone are not transmitted when their correlation is above a pre-defined threshold.

40. The device of claim 28, wherein the neck-wearable housing generally is U-shaped.

41. The device of claim 28, wherein the neck-wearable housing generally is O-shaped.

42. The device of claim 28, wherein the neck-wearable housing is flexible in at least one location.

43. A wearable telecommunication device, comprising: a flexible neck-worn sheath with at least one electronic unit attached thereto; two in-ear earphones; two cords, one of which connects to one of the in-ear earphone to the electronic unit, and the other cord connects the other in-ear earphone to the electronic unit, wherein the two cords are mechanically connected to the neck-worn sheath, and points of connection of the two cords to the neck-worn sheath are close to each other and form a dorsal cord connection node, and are further mechanically connected to each other in sections between the in-ear earphones and the dorsal cord connection node to form a suboccipital cord connection node at the connection point; and a rear microphone in the suboccipital cord connection node or in the dorsal cord connection node.

Description

BRIEF DESCRIPTION OF THE ATTACHED FIGURES

(1) The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain the principles of the invention.

(2) In the drawings:

(3) FIG. 1 shows a general view of a wearable device according to the invention in the operational position of the earphones, placed on the user.

(4) FIG. 2A and FIG. 2B illustrate a mathematical model of the neck surface where a headset has two nodes and the head is shown in the normal position and may rotate by an angle π/2.

(5) FIG. 3A and FIG. 3B illustrate a mathematical model of the neck surface where a headset has a single node and the head is shown in the normal position and may rotate by an angle π/2.

(6) FIG. 4A and FIG. 4B illustrate a mathematical model of the neck surface where a headset has two side nodes and the head is shown in normal position and may rotate by an angle π/2.

(7) FIG. 5A illustrates a mathematical model of the neck surface where a headset has two nodes and the head is tilted vertically.

(8) FIG. 5B illustrates the calculation of the length of segment AB when the head is tilted forward by an arbitrary angle α.

(9) FIG. 6A and FIG. 6B illustrate a mathematical model of the neck surface where a headset has a single node and two side nodes, correspondingly, and the head is tilted vertically.

(10) FIG. 7A and FIG. 7B illustrate a mathematical model of a head tilted sideway where a headset has two nodes and where a headset has two side nodes.

(11) FIG. 8 illustrates a mathematical model of a head tilted sideways where a headset has a single node.

(12) FIG. 9 shows dependence of the length of geodesic line AC on the head tilt angle α.

(13) FIG. 10 is a vector diagram of forces.

(14) FIG. 11 is a general view of a wearable device according to the invention, with neck-wearable housing having a generally U-shape, showing the primary functional components.

(15) FIG. 12A and FIG. 12B show an embodiment of a headset according to the invention, having a suboccipital node in the form of a clip.

(16) FIG. 13 shows embodiments of a wearable device according to the invention, wherein the cord portion between the suboccipital and dorsal cord connections nodes is provided in the form of a helical cord.

(17) FIG. 14 shows an embodiment of a headset according to the invention, where the integral cord portion between the suboccipital and dorsal cord connection nodes is provided in the form of an S-shaped spring.

(18) FIG. 15 shows embodiments of a headset comprising an electronic unit, according to the invention.

(19) FIG. 16 shows an example of electronic components incorporated into the electronic unit, according to the invention.

(20) FIG. 17 shows an embodiment of a neck-wearable housing connector according to the invention.

(21) FIG. 18 shows an embodiment of a wearable device according to the invention, with four microphones.

(22) FIG. 19 and FIG. 20 show illustrative circuitry of a headset embodiment according to the invention.

(23) FIG. 21A illustrates ideal propagation of a spherical acoustic wave front when a user speaks.

(24) FIG. 21B illustrates real propagation of a spherical acoustic wave front, taking in account an acoustic shadow area behind the speaking user.

(25) FIG. 22A, FIG. 22B and FIG. 22C illustrate equipment connection diagram and disposition of microphones used for experimental measurements of the invention prototypes.

(26) FIG. 24A shows a schematic diagram illustrating noise canceling in low frequency signals, according to the invention.

(27) FIG. 24B and FIG. 24C show schematic diagrams illustrating noise canceling in high frequency signals, according to the invention.

(28) FIG. 23A shows the difference between amplitude-frequency response curves of microphones placed on the chest, in the auricle, and on the dorsal neck surface of a user.

(29) FIG. 23B shows the difference between phase-frequency response curves of microphones placed on the chest, in the auricle, and on the dorsal neck surface of a user.

(30) FIG. 25 shows the position of a rear microphone placed in the suboccipital node of a wearable device according to the invention.

(31) FIG. 26A shows a schematic diagram of a Wiener adaptive filter according to the invention.

(32) FIG. 26B shows a schematic diagram of a least-mean-square method filter according to the invention.

(33) FIG. 27 shows a connection diagram of a microphone array comprising a rear microphone, in one embodiment of the invention.

(34) FIG. 28 illustrates positions of microphones of a microphone array comprising a rear microphone, in one embodiment of the invention.

(35) FIG. 29A illustrates possible positions of a rear microphone on the user's body in one embodiment of the invention.

(36) FIG. 29B illustrates possible positions of front microphones on the user's body in one embodiment of the invention.

(37) FIG. 30 illustrates possible positions of front microphones on the user's body in one embodiment of the invention comprising an additional device in a form of glasses.

(38) FIG. 31 illustrates possible positions of a rear microphone on the user's body in one embodiment of the invention comprising a dorsal node and a suboccipital node, when the user rotates his head

(39) FIG. 32 illustrates position of a rear microphone on the user's body in one embodiment of the invention, when a dorsal node is retracted and attached to a suboccipital node.

(40) FIG. 33 illustrates position of rear microphones on the user's body in one embodiment of the invention, when a dorsal node is detached from a suboccipital node and placed on the user's dorsal surface.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

(41) Reference will now be made in detail to the preferred embodiments of the present invention, examples of which are illustrated in the accompanying drawings.

(42) In order to assure wearing a device on the user's body without additional support, it is expedient to provide the device in the form of a loop or a half-loop.

(43) When a user wears a headset with neck-wearable housing having a generally U-shape (FIG. 1), a node that connects earphone cords 5 to a neck-wearable housing 1 rests on the dorsal surface of the user's neck, in the region of the seventh cervical vertebra. Slightly lower on the human body, there is a deepening trough, lying between the spinous and transverse processes of the vertebrae, sulcus dorsalis, at the level of the second—third thoracic vertebrae in the interscapular region, where a depression of various depth of about 4×5 cm (depending on the user's constitution and development of subcutaneous fat) is formed at the place of attachment on the medial edges of both blades of serratus anterior muscle, and a large and minor rhomboids muscles (musculae rhomboidei major et minor). The depression may receive a cord winding mechanism and an earphone storage pocket, without projecting above the surface of the body and so without causing inconvenience to the user.

(44) From the cord connection node on the neck part, the cords run up on the dorsal surface of the neck to the back of the head, on the paravertebral deepening, sulcus costae vertebralis major, not reaching the outside occipital protuberance at the level of the first or second cervical vertebrae, where an additional cord connection node, suboccipital node 6, is appropriate to arrange. If the cords are directed in a V manner from the suboccipital node in the oblique anterior-upward direction slightly above or at the hairline, which is almost coinciding with the upper occipital skull line, through the mastoid regions (regiones mastoideae) of the neck, above the mastoid processes, through the projection of ligamentum auriclere superior, which attaches the top part of the auricler cartilage to the squamous part of the temporal bone on the upper portion of the auricle between the front curl and tragus of the outer ear to a fixation point in the earphone 3 of the appropriate side.

(45) Then the stable position of the suboccipital cord connection node will be provided by the availability of fixing anatomical structures at the datum point, such as the external occipital protuberance and lateral occipital projections, while a snug fit of the cords on the scalp is provided by stretching them on the dorsal surface of the head and neck in the places where the cords pass like a girth due to the partial hook-like overlap of the earphone cords through the ligamentum auriclere with additional fixing of the earphones inside the auricle.

(46) With such attachment only the cords in the portion 7 between nodes 5 and 6 are movable, and only this portion may have a slack for compensation of the cord length, which changes when the head turns in the horizontal plane, tilts back, rocks from side to side, as well as when the movements are combined, that is, in all options that can arise in closed kinematic chains of the neck.

(47) Cords 4 are relatively snugly fitted to the scalp and fixed relative to the user's head, and their length does not vary with all of the above movements and varies so little that these variations can be neglected.

(48) Adherence and immobility of the cords 4 between the nodes are also promoted by the cellular connective tissue structure of the subcutaneous fat of the occipital region, a minor displacement of the skin in the area, the presence of Langer's lines running in the transverse direction in the skin, as well as the passage of the cord on a hollow of the postural cavity, the hook-like overlap of the cords and positioning the earphones in the outer ear.

(49) In conjunction with the suboccipital node, the tension and absence of slack are further provided by the design of the earphone, which is placed inside the auricle, in most cases, without an arc, but having a stiff part—an earphone arm attached to the earphone body lying in the outer ear and continued upward from the helical root on the ascending part of the helix to the ligamentum auriclere superior, the attachment point of the top of the auricle to the temporal bone.

(50) A flexible cord extends from the stiff arm, leaning over the above ligamentum auriclere superior at an angle of less than 45°, which contributes to the fact that the rigid arm of the earphone forms a lever, where at accidental tearing off of the earphone cords, that is, when the cords are pulled at down and back tension vector, the arising moment abuts the earphone against the tragus, thereby fixing the earphone between the tragus and the external auditory canal.

(51) It shall be noted that it is preferred to use in-ear earphones, which are fully or partially inserted into the external auditory canal, in comparison with those in-ear earphones, which are placed within the auricle and are not inserted, wholly or partially, into the external auditory canal.

(52) In terms of biomechanics, it shall be noted that movements of the head are described on the basis of closed kinematic patterns, and extrapolation of even fairly complex combinations of head movements to the fixation points can be considered in only one narrative category—as lengthening-shortening the cord portion between the dorsal cord connection node on the neck-wearable housing and the cord connection suboccipital node, which is almost stationary relative to the head and lies under the outer posterior occipital protuberance.

(53) To construct a closed kinematic model, a headset can be represented as having two basic parts and a movable connection thereof (FIG. 1).

(54) A first part (head part) is stationary relative to the user's head; it has two earphones 3, two earphone cords 4 enveloping the auricle from above, and a suboccipital node 6.

(55) A second part is stationary relative to the user's body; it has a neck part 1 and a cord connection node disposed on the neck-wearable housing on the dorsal surface of the neck, a dorsal node 5.

(56) As shown in FIG. 1, positions of the cord connection nodes has been chosen at reference numeral 5, point A and reference numeral 6, point B (FIG. 2A, FIG. 2B). In this case, the length of the free-hanging cord 7 in the portion between the nodes should be minimal.

(57) To determine the length of the AB portion, variations in the distance between points A and B as the head turns are to be considered. In this case, “distance” is the length of the geodesic line connecting points A and B on the surface of the neck (FIG. 2B). First, let us define the extension of the cord when the head rotates sideways. Maximum angle of rotation of the head is 90°. Let us determine the AB distance.

(58) To determine the length of the geodesic line it is necessary to describe mathematically the surface of the neck and possible movements of the head and neck. The neck surface can be represented with sufficient accuracy as a cylinder (FIG. 2A). Head and neck can make the following motions: bending-tilting forward, extension/tilting backward, abduction and adduction/tilting to the left and to the right, turns to the left and to the right. High movability of the cervical spine is caused by its segmentation: having a height of about 13 cm, it contains seven medium-sized vertebrae and six high intervertebral discs.

(59) Between the first cervical vertebrae and the occipital bone, in the atlantal-occipital joint, adduction/abduction and flexion/extension of the head are performed, and between the first and second cervical vertebra turns of the head to the right and the left are performed. The joint work of these joints provides the head movement about three axes. Thus, combined movements of the head and neck are made in relation to the body, while independent movements of the head are made in relation to the neck. This is because the cervical spine is very flexible, and independent movements are possible between the first and second cervical vertebrae.

(60) Let us consider the behavior of the kinematic model of the headset when the head rotates in the horizontal plane.

(61) When the head rotates in the horizontal plane, the neck twists mainly in the region between the first and second vertebrae. Moreover, since the cervical spine is located closer the back of the neck, the twisting axis is also close to the back surface of the cylinder. Since the twisting is performed only in the upper part of the cylinder about a non-central axis, the cylinder surface is distorted. The distortion is most strongly manifested in the region of the first and second cervical vertebrae, just where point B lies.

(62) The main part of the geodesic line passes below the distortion, so in the calculations we assume the surface as cylindrical. An important issue is the determination of the location of point B when the upper part of the cylinder is twisted to a maximum angle α=π/2. Since ears are symmetric about the twisting axis, that is the axis of the vertebral column, and the point B is fixed by the tensioned cords in symmetrical position as well, the position of point B can be expected in the next central angle φ (FIG. 2B).

(63) $\begin{array}{cc}\phi =\arcsin \left(\frac{R-D}{R}\right)& \left(1\right)\end{array}$

(64) The height of point B will not change at rotation either, because it is fixed by the tensioned earphone cords.

(65) Let us consider the task of geodesic line of a cylinder having base radius R and height h (FIG. 2B). The line passes through two diametrically opposite points on different basis.

(66) The geodesic line length is
ds=√{square root over (dx.sup.2+dy.sup.2+dz.sup.2)}

(67) shown in differential form.

(68) Since the curve lies on the surface of the cylinder, it is convenient to use cylindrical coordinates, with dx.sup.2+dy.sup.2=R.sup.2dφ.sup.2, where φ is the polar angle (FIG. 2B). In polar coordinates, the task is to find dependence z(φ), at which the length of the curve is minimal or when the function

(69) $\begin{array}{cc}S=\underset{0}{\overset{{\phi }_0}{\int }}\sqrt{R^2+z^{\mathrm{\prime 2}}}d\phi & \left(2\right)\end{array}$

(70) is minimal.

(71) From calculus, a minimum is reached for the curve that satisfies the Euler equation, in this case:

(72) $\begin{array}{cc}{\left(\frac{z^{\prime }}{\sqrt{R^2+z^{\mathrm{\prime 2}}}}\right)}^{\prime }=0& \left(3\right)\end{array}$

(73) It follows that z′(φ)=a, where a is the constant factor, then z(φ)=a×φ+b. Coefficients are determined though boundary points A (R,0,0), the attachment point of the lower clip, and B (R,φ0,h) with the polar angle φ=0 being at point A and equal to φ.sub.0 at point B. Then the coefficients are of the form: a=h/φ.sub.0, b=0. Then z(y)=φ×h/φ.sub.0. And the length of the curve is equal to the value of the functional, i.e.:

(74) $\begin{array}{cc}S=\underset{0}{\overset{{\phi }_0}{\int }}\sqrt{R^2+h^2/{\phi }_0^2}d\phi =\sqrt{{\phi }_0^2{R}^2+h^2}& \left(4\right)\end{array}$

(75) Thus, variation in distance AB or movability of cords ΔS is:
ΔS=√{square root over (h.sup.2+R.sup.2φ.sub.0.sup.2)}−h  (5)

(76) where R—the radius of the cylinder, φ.sub.0—the angle of rotation of node B, defined relative to the central axis of the cylinder, h—the height of the node. With regard to expression (1) the expression for movability of the cords is:

(77) $\begin{array}{cc}\Delta S_t=\sqrt{h^2+R^2{\arcsin }^2\left(\frac{R-D}{R}\right)}-h.& \left(6\right)\end{array}$

(78) Now, for comparison, we will consider variation in the length of cords at horizontal rotation of the head in conventional headsets. FIG. 3A shows an example of such a headset. In this case, cords are clamped at point A, and the movable part is the entire cord from point A to earphones disposed at points C and D.

(79) Conventionally, the headset is denoted as a single node headset. Thus, movability of the cords can be determined from the difference between the distances from point A and D when the head rotates at the angle of 90° in one direction and in the other direction, since while the distance or the geodesic line length increases in one direction, it decreases in the other direction. These two distances can be determined in FIG. 3B, where the minimal distance is the length of line AC, and the maximum distance corresponds to line AD. As a result, movability of the cords can be found from expression (5) with the assumption of h=H and φ.sub.0=π, and it has the form:
ΔS.sub.t1=√{square root over (H.sup.2+R.sup.2π.sup.2)}−H  (7).

(80) Let us consider another type of a headset, which will be conventionally called a headset having two side nodes (FIG. 4A). In this case assume that the headset cord, at rotation, always passes through points at the base of the cylinder, i.e., points A and B, cord connection nodes. Then the minimum distance between points A and C or B and D is H. The maximum distance when the head is rotated to 90° is shown by geodesic lines AC and BD (FIG. 4B). As a result, movability of the cords is also determined from expression (5) with the assumption of h=H and φ.sub.0=π/2, and is defined by the following expression:
ΔS.sub.t2=√{square root over (H.sup.2+R.sup.2π.sup.2/4)}−H  (8).

(81) Next, let us consider behavior of the kinematic model when the head tilts forward and backward in the vertical plane.

(82) Tilts of the head are performed by rotation of the head around the axis extending between the first cervical vertebra and the occipital bone. The tilt is often accompanied by a tilt of the entire cervical spine. In a headset having two nodes, the tilt of the neck has a little effect on distance AB, but rotation of the head has a significant impact, since node B is disposed directly on the occipital part. Thus, knowing distance from B to axis of rotation r and angle of rotation α (FIG. 5A), shift of node B can be estimated as
BB.sub.0=rα  (9).

(83) Now we will obtain an expression for the length of segment AB at arbitrary angle α from the triangle AOB (FIG. 5B):
AB.sup.2=AO.sup.2+r.sup.2−2AO×r×cos(α+β)  (10).

(84) Distance to axis r can be determined though the distance from the back surface of the neck to the center of the cervical spine, i.e., R-D, and the difference of heights of point B and the axis of rotation of the head h.sub.0:
r=√{square root over ((R−D).sup.2+h.sub.0.sup.2)}  (11).

(85) Then we will obtain the following expression from triangle OO.sub.1A:
AO=√{square root over ((R−D).sup.2+(h+h.sub.0).sup.2)}  (12).

(86) Expression for angle β can be obtained from expressions (10), (11) and (12) by substituting α=0, AB=h.

(87) $\begin{array}{cc}\beta =\arccos \frac{{\left(R-D\right)}^2+{\mathrm{hh}}_0+h_0^2}{\sqrt{\left({\left(R-D\right)}^2+{\left(h+h_0\right)}^2\right)\left({\left(R-D\right)}^2+h_0^2\right)}}.& \left(13\right)\end{array}$

(88) Thus, the expression for AB has the form:

(89) $\begin{array}{cc}\mathrm{AB}\left(\alpha \right)=\sqrt{2{\left(R-D\right)}^2+{\left(h+h_0\right)}^2+h_0^2-2\sqrt{\left({\left(R-D\right)}^2+{\left(h+h_0\right)}^2\right)\left({\left(R-D\right)}^2+h_0^2\right)}\cos \left(\alpha +\beta \right)}.& \left(14\right)\end{array}$

(90) In case of tilting, the head backward expression (14) is no longer true, because there is no tension of the skin and soft tissues of the dorsal part of the neck. In this case it is appropriate to estimate distance BB0 as the difference between heights of points B and B.sub.0:
Δh=r(cos(γ.sub.0+α)−cos γ.sub.0)  (15).

(91) As a result, movability of the cords is calculated from expression (14) by substituting α=α.sub.m (maximum tilt angle), and (15) by substituting α=−α.sub.m:
ΔS.sub.c=AB(α.sub.m)−√{square root over ((R−D).sup.2+h.sub.0.sup.2)}(cos(γ.sub.0−α.sub.m)−cos γ.sub.0)  (16).

(92) Apparently, α.sub.m cannot exceed γ.sub.0 due to the limit on deformation of the neck. To assess movability of the cords, we may assume α.sub.m=γ.sub.0, then with regard to expression (14) we may obtain:
ΔS.sub.c=AB(γ.sub.0)−√{square root over ((R−D).sup.2+h.sub.0.sup.2)}(1−cos γ.sub.0)  (17).

(93) In case of headsets with a single node or with two side nodes rotation in the vertical plane affects the height of points C and D. Variation in the latter, Δh.sub.0, can be determined if relative distance r.sub.0 between axis CD and the axis of rotation, as well as angular position α.sub.0 of the axes are known (FIG. 6A):
Δh.sub.0=r.sub.0(cos α.sub.0−cos(α.sub.0+α))  (18).

(94) As a result, variation in the distance or movability of cords for a headset having a single node can be obtained from expression (4) with H−Δh.sub.0 set instead of h and φ=π/2. In this case, angle α varies in the range −α.sub.m<α<α.sub.m, and the height varies in the range:
Δh.sub.01=r.sub.0(cos α.sub.0−cos(α.sub.0−α.sub.m))<Δh.sub.0<r.sub.0(cos α.sub.0−cos(α.sub.0+α.sub.m))=Δh.sub.02  (19),
ΔS.sub.c1=√{square root over ((H−Δh.sub.01).sup.2+R.sup.2π.sup.2/4)}−√{square root over ((H−Δh.sub.02).sup.2+R.sup.2π.sup.2/4)}  (20).

(95) FIG. 6B illustrates the case of a headset having two side nodes. Movability of the cords can be estimated through variation in heights of points C and D. Then, from expression (19) we can obtain movability of the cords in the following form:
ΔS.sub.c1=Δh.sub.02−Δh.sub.01  (21).

(96) Like in the case of a headset having two nodes, estimates α.sub.m=γ.sub.0=α.sub.0 are true. Then we may obtain the following estimate for movability of cords:
ΔS.sub.c1=√{square root over ((H+r.sub.0(1−cos γ.sub.0)).sup.2R.sup.2π.sup.2/4)}−√{square root over ((H−r.sub.0(cos γ.sub.0−cos 2γ.sub.0)).sup.2+R.sup.2π.sup.2/4)}  (22),
ΔS.sub.s2=r.sub.0(1−cos 2γ.sub.0)  (23).

(97) Also consider behavior of the kinematic model when the head tilts sideway in the vertical plane.

(98) When the head tilts sideway, the movement of the head can be represented as rotation of the upper part of a cylinder about axis s, which extends approximately through point O of intersection of axes t and c.

(99) In the case of a headset having two nodes, such rotation is accompanied by a shift of point B, which can be estimated through the distance to axis of rotation O.sub.1B.sub.0 (FIG. 7B). As seen in FIG. 7B: O.sub.1B.sub.0=h.sub.0. To determine the length of AB it is necessary to determine horizontal shift Δs and vertical shift Δh of point B, because AB=√{square root over ((h+Δh).sup.2+Δs.sup.2)}. In this case Δh=h.sub.0(1−cos α) and Δs=h.sub.0 sin α. Then movability of section AB when the head tilts sideway, will be changed to maximum angle α.sub.m:
ΔS.sub.s=√{square root over ((h+h.sub.0(1−cos α.sub.m)).sup.2+h.sub.0.sup.2 sin.sup.2α.sub.m)}−h  (24).

(100) Now let us consider the case of a headset having side nodes. In this case, variation in segments AC and BD can be accounted for by considering the shift of points C and D on arcs of circle from points C.sub.0 and D.sub.0. The length of AC in the case of the head tilt shown in FIG. 7B can be found as:
AC=AC.sub.0+R.sub.sα=H+R.sub.sα  (25).

(101) Here R.sub.s is the radius of rotation path about axis s, which can be found from triangle COO.sub.2, where OO.sub.2 can be found, given that the height of point O is h+h.sub.0 (FIG. 5B), then OO.sub.2=H−h−h.sub.0, and CO.sub.2=R, therefore:
CO=R.sub.s=√{square root over ((H−h−h.sub.0).sup.2+R.sup.2)}  (26).

(102) To determine BD, only variation in the height of point D, ΔH=R.sub.s sin α, should be taken into account because the cord in this area is loose:
BD=H−ΔH=H−R.sub.s sin α  (27).

(103) Considering maximum deflection angle α.sub.m=45°, the following expression can be obtained for movability of cords:
ΔS.sub.s2=R.sub.sα.sub.m+R.sub.s sin α.sub.m  (28).

(104) Now we will consider the case of a headset having a single node (FIG. 8). In this case, the calculation is more complicated and requires special treatment for the length of geodesic line AC. In this task the surface of the neck can be described as a surface of an inclined cylinder. To do this, let us find the angle of inclination of the cylinder surface, β. From triangles BCC.sub.0 and OCC.sub.0 we may find CC.sub.0=2R.sub.s sin(α/2):
BC=√{square root over (H.sup.2+4R.sub.s.sup.2 sin.sup.2(α/2)−4HR.sub.s sin(α/2)sin(α/2−γ))}  (29).

(105) From triangle BCC.sub.0 we may obtain:
BC/sin(π/2−a/2+γ)=2R.sub.s sin(α/2)/sin β
so we may obtain:
β=arcsin(2R.sub.s sin(α/2)cos(α/2−γ)/BC)  (30).
Here
γ=arctan(R/(H−h−h.sub.0))  (31).
Therefore,
AC=√{square root over ((BC(1−sin β)).sup.2+π.sup.2R.sup.2 cos.sup.2β/4)}  (32).

(106) It should be noted that, taking into account the dependence of BC and β on angle α from equations (29) and (30), we can expect a non-monotonic dependence of the line length AC(α). FIG. 9 shows this dependence for parameters listed in Table 1. It can be seen that AC reaches maximum AC.sub.max=16.6 cm at angle α.sub.0=8.6°.

(107) Now let us find the length of AD as this line describes the minimum length of the cord. In this case we may consider that the height of the cylinder has changed to ΔH=R.sub.s sin α, then using the expression (27) we may obtain:
AD=√{square root over ((H−R.sub.s sin α).sup.2+π.sup.2R.sup.2/4)}  (33).

(108) As a result, movability of cords ΔS.sub.s1 is determined as the difference of the lengths of lines AC.sub.max and AD at the maximum angle of inclination, α.sub.m:
ΔS.sub.s1=AC.sub.max−√{square root over ((H−R sin α.sub.m).sup.2+π.sup.2R.sup.2/4)}  (34).

(109) Table 1 shows the comparison of cord movability for various types of headsets. As seen in the table, a headset having two nodes, that is, a headset in which two earphone cords are connected to the neck-wearable housing through a dorsal cord connection node in close proximity to each other and have an additional point of fixation to each other, i.e., a suboccipital node; the cords have the lowest movability as compared with conventional headsets. This advantage applies to all kinds of movements of the head.

(110) Comfortable wear of the headset is determined by the maximum possible movability of cords, respectively, the difference between the minimum and maximum possible length of a loose cord, arising at different positions of the head. In a headset having two nodes, the maximum length is determined by maximum distance AB between the nodes, that is, the length AB defined in expression (14). In a headset having a single node, the maximum length of the cord is achieved when the head rotates to 90°:
L.sub.max 1=√{square root over (H.sup.2+R.sup.2π.sup.2)}  (35).

(111) For a headset having two side nodes we may obtain the maximum length when the head tilts sideway:
L.sub.max 2=H+R.sub.sα.sub.m  (36).

(112) Table 1 contains numerical estimates, from which it follows that the headset having two nodes has a minimum length of a maximum extended, but slack portion of cord. It should also be noted that the estimates obtained for a headset having two side nodes have been deliberately reduced because cords passing from points A and B to the transceiver are not taken into account, and account of them would significantly increase L.sub.max2.

(113) Therefore, the availability of two optimally positioned nodes A and B contributes not only to reduction in slacking of the cords, but also provides tension of the cords extending from node B to earphones. Since these cords lie on the curved surface of the neck, the tension creates a pressure on the skin (FIG. 10). As a result of this pressure there arises a friction force of the cord against the skin and a pressure force of the suboccipital cord connection node, node B, against soft tissues, while the difference of vectors of these forces leads to fixation of the cords on the scalp and further secures earphones in the auricle. Thus, the securing force is concentrated not only on the auricles, and not only by fixing the earphones in the external auditory canal, but it is uniformly distributed over the entire length of the cord, which greatly facilitates wearing of the earphones. Node B, i.e., suboccipital cord connection node, is held in a stable position owing to the uniform distribution of various forces that arise in the occipital region at the specified arrangement geometry of the cords and their mutual coupling, taking into account human anatomical features.

(114) FIG. 10 shows a vector diagram of projections of the forces acting on the suboccipital cord connection node, node B. Node B is fixed through tension of the cords. Thin arrows indicate tensile forces of the cords, the total of which creates pressure on the skin. As a result, node B undergoes a force of reaction of the skin and surrounding tissues, indicated by wide hollow arrow, that seeks to move the node down, and the arising forces of friction against the cord, marked with wide solid arrow, fix the position of node B. In this model, the tension of cords below the node was neglected, as its length has been chosen for optimal and the cord is loose, and it has an excess length of about 9.8 cm to ensure movability of the cords in movements of the head and neck.

(115) Table 1 summarizes results of comparison of cord movability and maximum cord length in headsets with different geometries.

(116) TABLE-US-00001 TABLE 1 Comparison of movability and maximum length of cord in headsets with various geometry Movement Rotation of head Tilt of head Tilt of head in horizontal plane forward/backward sideway Estimate at Estimate at Estimate at Headset type Expression parameters Expression parameters Expression parameters Headset with (6) R = 6.5 cm, (12) R = 6.5 cm, (24) R = 6.5 cm, two nodes h = 6 cm, h.sub.0 = 2 cm, h = 6 cm, D = 1 cm, D = 1 cm h.sub.0 = 2 cm α = 90° = 1.6 rad ΔS.sub.c = 8.6 cm α.sub.m = 45° = 0.8 rad ΔS.sub.t = 2.9 cm ΔS.sub.s = 0.6 cm L.sub.max = 9.8 cm (see expression (14)) Headset with (7) R = 6.5 cm, (18) r.sub.0 = 3 cm, (34) R = 6.5 cm, a single node H = 13 cm γ.sub.0 = 45° h = 6 cm ΔS.sub.t1 = 12.5 cm ΔS.sub.c1 = 2.2 cm h.sub.0 = 2 cm α.sub.m = 45° = 0.8 rad ΔS.sub.s1 = 3.4 cm L.sub.max 1 = {square root over (H.sup.2 + R.sup.2π.sup.2)} = 25.5 cm Headset with (8) R = 7 cm, (19) r.sub.0 = 3 cm, (28) R = 6.5 cm, two side nodes H = 13 cm γ.sub.0 = 45° h = 6 cm ΔS.sub.t2 = 5.2 cm ΔS.sub.c2 = 3 cm h.sub.0 = 2 cm α.sub.m = 45° = 0.8 rad ΔS.sub.s2 = 11 cm L.sub.max 2 = H + R.sub.sα.sub.m = 19.4 cm

(117) The technical effect provided by the invention includes the ability to reduce the length of the movable portion of the cords between the earphone and the neck-wearable housing, and the adherence of the stationary portion of the cord to the surface of the user's body and fixation of the stationary portion by tension, to substantially eliminate slack of the cords connecting the earphones with the neck-wearable housing, which in turn, prevents breakage of cords or earphones, and provides an additional opportunity for constant wear of the headset by the user in the operational position or with the earphones taken off, because the cords do not impair the aesthetic appearance of the user when the earphones are worn in the operational or non-operational position. Furthermore, a mechanism for full or partial winding up the earphone cords when not in use can be arranged on the headset more easily.

(118) A headset for a mobile electronic device (FIG. 1, FIG. 11) comprises a neck-wearable housing 1 having a generally U-shape with at least one electrical connector 2 attached to it; two in-ear earphones 3A, 3B; two cords 4A, 4B, each connected at one end to a corresponding in-ear earphone and connected at its other end to the electrical connector and these the two cords are mechanically connected to the neck-wearable housing 1, and points of connection of the cords to the neck-wearable housing are in close proximity to each other and form a dorsal cord connection node 5. They are also mechanically connected to each other in sections between the in-ear earphones and the dorsal cord connection node 5 to form a suboccipital cord connection node 6 at the connection point.

(119) When the headset is worn in the operational position, the dorsal cord connection node 5 and the suboccipital cord connection node 6 are located on the dorsal surface of the neck, and cords 4 in sections 7 between the earphones 3 and the suboccipital node 6 are located over an auricle.

(120) In various embodiments (FIGS. 12A, 12B) the suboccipital cord connection node may be a clip 8, and the length of the cords can be adjusted by moving the clip along the cords. Also, the suboccipital node may include an electrical connector to disconnect the cords.

(121) At least one cord in the section 7 between the suboccipital and dorsal nodes can be configured as a helical spring. In the embodiment shown in FIG. 13, the section 7 between the suboccipital and dorsal cord connection nodes is configured in the form of a helical cord and acts as a tension spring. The cords may be formed not only in the form of a helical spring, but also in the form of a flat S-shaped spring, which may be more expedient in some cases, as this allows avoiding distortion of the spring and provides a precisely adjacent position of the suboccipital node in relation to the device when the earphone is in the non-operational position.

(122) An embodiment of the headset using an S-shaped spring is shown in FIG. 14.

(123) In preferred embodiment (FIG. 15), a headset comprises at least one electronic control unit 9 mechanically and electrically connected to an electrical connector. The electronic unit 9 can be placed on the chest at one side of a neck-wearable housing. The electronic unit 9 may comprise a signal transceiver, such as Bluetooth, to receive signal from a cell phone; there may further be a battery, a player, a radio, a USB flash drive, an electronic key, a satellite signal receiver such as GPS and/or GLONASS receiver able to tell coordinates through voice information transmitted directly to the user's earphones.

(124) FIG. 16 shows possible electronic components incorporated into the electronic unit 9. The electronic unit 9 communicates with a mobile phone, a satellite navigation system, a computer or a mobile station via a radio communications module 10. A signal processor 11 processes audio signals, and controls and manages data streams. Digital-to-analog conversion and amplification of a signal, and volume control are performed by a codec or an audio module 12. A memory module 13 stores control software, hardware setting profiles and user's information. A power source, such as a battery 14 incorporated in the electronic unit 9 and/or disposed on the neck-wearable housing, provides operation of microcircuit chips. The electronic unit 9 may include control buttons, such as 15, 16, and 17. A short-range Near Field Connection module 18 can be used for data exchange and quick coupling with a mobile electronic device.

(125) In various embodiments of a headset the electronic unit 9 accommodates the following accessories: an extra controller 19 for processing signals from control buttons; a slot 20 with a connector to connect an external flash memory, a USB connector 21 for data transfer or charging the battery. Connectors 22 are used to connect earphones, external microphones, and additional control buttons.

(126) Buttons 15, 16 can be disposed on the neck-wearable housing 1 (FIG. 15).

(127) In various embodiments of the headset, control buttons and keys are disposed both on the housing of the electronic unit 9 and the neck-wearable housing.

(128) Furthermore, pressure can be made at once on two opposite buttons with two fingers, thumb and forefinger, simultaneously on both sides of the neck-wearable housing relative to the electronic control unit or the rigid member disposing on the neck loop. This eliminates accidental pressure by a vehicle safety belt, a bag strap, etc. Such an arrangement of buttons provides for maximum accessibility to them, even when wearing a tie, suit or coat.

(129) The headset (FIG. 15) may further comprise a rigid member 23 are configured like shells having a polyhedron-like shape, wherein a planes of two narrow facets of the polyhedron-like shape are substantially parallel to each other and substantially perpendicular to the user's body plane, and the rigid member contains a button 17 positioned on the facets near the rigid member's corner where the facet borders another facet being perpendicular to the user's body plane, and the button is configured to be easily found and activated through the clothes when the device is worn under the clothes, wherein intentional activation thereof is possible by catching and holding the member with two fingers.

(130) Herein, control means of the claimed device are mainly described as buttons or keys in examples and embodiments. However, other types of control means may be used depending on the functions controlled by these control means.

(131) In an embodiment of the headset (FIG. 15), a power supply 24 and a microphone 25 can be disposed on a neck-wearable housing 1, for example in a rigid member 23. The power source (e.g., batteries) may be distributed in the electronic unit 9.

(132) In an embodiment of the headset (FIG. 18), four microphones 17A-D can be disposed on a neck-wearable housing 1 forming a microphone array.

(133) In some embodiments, the headset can be free of cords transmitting signal to the earphone and have a power cord only; a cordless module in each earphone to receive and transmit electromagnetic signal for the earphone.

(134) A neck-wearable housing 1 (FIG. 17) may have at least two slots 26 for connecting additional sections of the neck loop.

(135) FIGS. 19, 20 show a connection circuitry of keys and microphones in one embodiment of the headset. This embodiment includes twelve keys and seven microphones.

(136) Data outputs to contact members, sync signal inputs of the microphones and control keys are connected to inputs of a signal processor or controller 27 (FIG. 20). Earphones are connected to a control chip, CODEC 28, or an audio module, which comprises a digital-to-analog converter, and a controllable-gain amplifier. In operation, the processor data exchanges data with peripheral devices 29 as well. FIG. 19 shows electrical circuitry of the headset.

(137) Unlike known stereo Bluetooth headsets, the electronic necklace in the form of an open or a closed loop can be controlled directly through the clothes, with no necessity of pulling it from a pocket of a bag or drawing it from under the clothes.

(138) The headset may also comprise gyroscopes, accelerometers, magnetometers or other position sensors to assist in navigation with voice prompts of GPS device.

(139) Benefits of the invention including: shortening by more than two times the length of the movable parts of cords, i.e., the portions between the nodes; convenient position and tension of cords on surface of the body; and immobility of the remaining cord portions allow the headset to be worn under clothes in the operational and non-operational position, and throat microphones may be disposed thereon.

(140) In many embodiments the headset can be controlled without taking it from under the clothes or pulling a phone from a pocket, because the buttons located under clothes can be pressed from outside, over clothes, or by giving voice commands without hand manipulations at all. Direct contact between the device and the user's skin allows positioning on the headset sensors for monitoring the state of user's health, such as temperature, blood pressure, sugar, alcohol in skin secretions, etc., to monitor galvanic skin response for the purpose of control of the sympathetic nervous system, which allows using the headset as a part of a biotelemetry system for medical diagnostics.

(141) The headset can be used not only as an option for connecting to a mobile phone or itself used as a mobile phone, but also as a component of a wearable mobile system with hardware distributed over several devices carried by a person, for example, some of hardware and battery base can be accommodated in a man's trouser belt, while the wired connection to the headset can be implemented in a cord, which lies under the clothes along the user's spine on the back; the headset itself can implement functionality of a mobile phone or smartphone, while a separately worn screen/keyboard unit can be used as a wireless interface to the mobile phone or smartphone.

(142) The headset design comprising a suboccipital cord connection node and a short, as compared to the other neck headsets, section of the movable portion of cords connecting the headset with the neck-worn housing allows wearing the headset under user's clothes, thereby eliminating the use of external microphone close to the user's mouth. This leads to the need to provide a special arrangement topology of microphones in the headset and a hardware/software system for processing signals from microphones.

(143) The problem of noise reduction in speech signals became even more pressing as the wearable devices may be used in a very noisy environment.

(144) Speech of a human being is a mix of tones with a lot of harmonic components and various noises, which mix very fast changes in time. The most representative frequency range of the human speech is 250 to 3000 Hz. The wave front of the sound is commonly spherical with the center of the sphere located at the speaker's mouth (FIG. 21A). The waves tend to propagate in all directions; however the real propagation model depends on the acoustic environment and the diffraction and interference pattern which in turn depends on the wavelength.

(145) The waves corresponding to speech components over 2000 Hz are almost unable to pass around the human head (which dimension is about 21 cm), so a so-called “acoustic shadow” is formed in the area of the occipital surface and the dorsal neck surface of the user (FIG. 21B). Lower frequency sounds below 340 Hz having the wavelength of 1 m and more are able to pass around the human head and reach the occiput area at the cost of some amplitude decrease.

(146) An anthropometric dummy was used in the acoustic field research performed by the inventor. FIG. 22A illustrates equipment connection diagram, where the test setup comprises an artificial voice unit 30 (Bruel & Kjaer, type 4128) connected to a power amplifier 31 (Bruel & Kjaer, type 2636), a number of microelectromechanical system (MEMS) microphones 33-38 (Knowles, type SPH1642HT5H-1) connected to a preamplifier 32 (Bruel & Kjaer, type 1054), and a measurement unit 2 (Audiomatica, CLIO) connected to a computer. FIG. 22B and FIG. 22C illustrate disposition of the microphones on the dummy's surface.

(147) A variable frequency sinusoidal signal was supplied from the measurement unit 39 via the power amplifier 31 to the artificial voice unit 30. The microphones 33-38 were alternately connected to the measurement unit 39 via the preamplifier 32. A quiet room was used for the measurements and a gain-frequency characteristic and a gain-phase characteristic were determined for each of the microphones.

(148) Irregularity of the obtained gain-frequency characteristics is mostly caused by frequency features of the artificial voice unit. As relative characteristics measured in the target points were in the experiment focus, the irregularity may be neglected. The front microphone 36 was used as a main microphone, and an additional reference microphone (not shown in the figures) disposed in front of the dummy's mouth was used in order to determine the relative characteristics of the microphones 33-38.

(149) The experiments yielded the following results (FIG. 23A, FIG. 23B): the directional diagram of the artificial voice unit 30 was found mostly even in the frequency range below 2 kHz and substantially uneven in the frequency range above 2 kHz, while the most sharp unevenness was found in the region of 10 kHz and more; the most difference (20 dB) in the frequency range above 2 kHz was found between the gain-frequency characteristics of the main microphone 36 centrally positioned on the dummy's chest and the rear microphones 37, 38; a moderate difference of 5 to 10 dB in the frequency range above 2 kHz was found between the gain-frequency characteristics of the microphone 33 positioned near the dummy's auricles and the rear microphones 37, 38.

(150) It is known that increasing signal-to-noise ratio (SNR) just by 1 dB is able to improve the speech intelligibility, which is usually evident to experts during experiments and being beyond possible experimental error range.

(151) Thus, the main microphone 36 positioned on the dummy's chest provides maximum level of a speech signal comparable with the reference microphone level, while the microphones 37, 38 positioned on the dummy's occiput or dorsal neck area provides minimal level of the speech signal in the frequency range above 2 kHz. This frequency range is particularly important for correct transmission of the speech via communications channels, as harmonic components providing the voice personalization mostly occupy the range over 1.7 kHz. Therefore, a pair of microphones, one of which is located on the user's chest and the other one is located on the back part of the user's neck, may be used for separation of the target speech signal.

(152) This is apparent from the above data, that positioning a microphone in close vicinity of the user's auricle (like the microphone 33), provides substantially less speech amplitude difference between this side microphone and the front microphone (like the main microphone 36), than the difference between a rear microphone (like the microphones 37, 38) and the same front microphone.

(153) Most of the above-mentioned conventional solutions use a front microphone located close to the user's mouth (like a boom microphone), so the speech amplitude difference is greater than indicated in the above experimental data. However, when a microphone is aggregated with an earphone, some special solutions for preventing acoustic feedback between the microphone and the earphone have to be used, which makes the digital signal processing algorithm for noise reduction more complicated and less efficient.

(154) Therefore, the inventor worked on finding an optimal number and layout of microphones applicable to a neck-worn hands-free device and selecting the best processing method for noise reduction.

(155) Different noise reduction systems are used for increasing SNR of speech signals. The general principle of these systems is determination of a noise estimation for a source additive signal by mathematical methods, and further subtraction of the noise estimation from the additive signal. Functions of noise reduction algorithms applicable to a microphone array are shown in FIG. 24A and FIG. 24B, wherein dedicated microphones for receiving noise are used.

(156) One possible method is illustrated by FIG. 24A, wherein the noise signal u.sub.n obtained from a dedicated noise microphone is subtracted from signals u.sub.1 . . . u.sub.m obtained from another microphones located in the area of direct propagation of a speech sound wave. Further, the output signal u.sub.s-n is formed by the unit 39 from the subtraction results.

(157) In another method illustrated by FIG. 24B, spectrum-sensitive subtraction is performed, which takes into account the phase difference of the microphones, which difference is significant for high frequency signals. In this method, signals of all of the microphones are Fourier-transformed, afterwards the noise amplitude n is subtracted from signal amplitudes s.sub.1, . . . s.sub.m and the output signal u.sub.s-n is formed by the unit 39 from the subtraction results. The algorithm implies dividing the working bandwidth into a number of sub-bands and using individual weight factors for the noise signal u.sub.n and for the processed signals u.sub.1 . . . u.sub.m in each sub-band in order to form the output signal u.sub.s-n. Different embodiments of this method may be done like those disclosed in references [6], [7]. For instance, Fast Fourier Transformation (FFT) and correlation matrices may be used, like in a Wiener filter.

(158) The above methods are applicable when the microphones are located in the area of direct sound wave propagation, i.e., in an area where the wave front is ideally spherical.

(159) FIG. 24C shows a simplified embodiment of the method of FIG. 24B, wherein one of two front microphones is used as a noise microphone, and the other one is used as a main microphone. The unit 39 selects the best microphone and controls the switch 40, where the target signal level of the best microphone is higher by a predetermined threshold 41.

(160) The noise sources are usually distantly located so the distance between them and the microphones is far greater than the sound wavelength. Therefore, the acoustic pressure S in the wave is fairly defined by the spherical wave formula:

(161) $\begin{array}{cc}s=\frac{A\exp \left(-i{k}_s\left(r-\mathrm{vt}\right)\right)}{r^2},& \left(37\right)\end{array}$

(162) where A is the oscillation amplitude of the source, k.sub.s is the wave vector, v is the wave speed, r is the sphere radius, t is the time of propagation.

(163) It can be seen from the above expression that if the user's head dimensions are substantially less than the distance r, the wave may be considered flat and the amplitude may be considered constant. In this case, the noise signal amplitude is about the same for each of the microphones disposed in a neck-worn device. Moreover, in a room, the noise sound waves may be highly reverberated so the noise sound wave front may be gravely distorted. This means that the noise signal level of all the microphones is substantially equal, no matters which side a microphone is directed to. So it may be enough to simply subtract the signal of the rear microphone used as a noise microphone from the signal of the front microphone used as a main microphone. If a microphone array is used, the subtraction may be performed for signals u.sub.1 . . . u.sub.m obtained from all microphones except for the noise microphone (FIG. 24A).

(164) However, such a simple subtraction method does not take in account the phase incursion between the noise microphone and the other microphones. This difference may be substantial for high frequencies, when the distance between the microphones is comparable to the sound wavelength. In this case, the spectrum-sensitive subtraction may be used as shown in FIG. 24B.

(165) When possible, one or several front microphones may be positioned in a device worn on the user's head (like a helmet, glasses, massive headphones, etc.). If not, it is expedient to locate a front microphone in a wearable neckband, so that the microphone is positioned in the area of the joining point of clavicle and episternum. A couple of front microphones (a left front microphone and a right front microphone) located on the user's chest can be used in order to compensate the user's head rotation during the talk. In particular, signal phase and amplitude differences tend to occur due to the user's head tilt and rotation, as the distance between the user's mouth and the front microphones continuously varies and a kind of a comb filter is formed.

(166) A microphone signal adder may be used for forming a single front sound signal by addition of the signals obtained from the front microphones. Additionally, a dynamic range compressor (DRC) may be used for reducing the excessively wide dynamic range. These solutions have been described in the prior art, so their details are omitted for the sake of brevity. It just shall be noted that these solutions are quite rarely used in wearable devices, as they require a hardware having substantial dimensions, weight and power consumption, whereas the necklace form-factor easily allows using such solutions.

(167) The controlled switch solution of FIG. 24C is also applicable for selecting one of the two front microphones, depending on which side the user's head is rotated and/or tilted is. Preferably, directional microphones may be used as the front microphones, wherein the directional diagram of each microphone is headed up to the user's mouth. A dedicated phased microphone array or a gradient microphone array may be used instead of the directional microphones, which may be e.g., a couple of front microphones positioned on the central line, comprising an upper microphone placed closer to the user's mouth and a lower microphone placed farther to the user's mouth).

(168) The microphones shall be relatively fixed on the wearable device. A Bowden cable may be used for connecting members of the wearable device in order to prevent twisting thereof and flipping over the microphones. Rustle in the microphones may be reduced (not eliminated though) by using very smooth cases for the wearable device members and by noise insulation of the microphones within the cases. The microphone holes shall be placed in those faces of the wearable device members, which do not normally contact the user's clothes.

(169) The position of the noise microphone is stipulated by the pattern of acoustic field formed during the user's speech. In particular, the speech wave level is expected to be substantially lower near the back side of the user's neck (FIG. 21B) due to diffraction effect and the wave is expected to be delayed. This assumption is confirmed by the above-mentioned experimental data.

(170) Moreover, as can be seen in FIG. 23A, the gain-frequency characteristics are non-monotonic. This may occur due to acoustic wave interference and/or resonance effects in the sound source and the environment. However, as the general pattern of the gain-frequency characteristics is substantially the same, unevenness of the characteristics may be neglected and the signal level differences of different microphones may be considered. In particular, using the microphones 37, 38 as the rear microphones is preferable, because the microphones 37, 38 provide less level of the user's speech signal than the microphone 33 and facilitate more distinct separation of the environmental noise signal.

(171) In order to more clearly define the impact of the acoustic wave diffraction, the phase-frequency characteristics of the microphones shall be analyzed. The phase difference of sound oscillations for two microphones is the product of the wave vector k.sub.s and the distance l between the microphones:
Δφ=k.sub.sl  (38).

(172) If the acoustic wave interference is neglected, then the acoustic wave dispersion principle is:
ω=vk.sub.s  (39),

(173) where v is the wave speed. Combination of (38) and (39) yields the expression for the phase difference:
Δφ=ωl/v  (40).

(174) Thus, a linear dependence for the phase difference of sound oscillations for two microphones can be expected with the above assumptions applicable for low frequency. However, these assumptions are generally not valid in the frequency range over 2 kHz. This is illustrated in FIG. 23B, where the phase-frequency characteristic of the main front microphone 36 is nearly linear, while the phase-frequency characteristic of the rear microphone 37, 38 is substantially non-linear, and the linearity measure of the phase-frequency characteristic of the side microphone 33 is intermediate. Linearity of the phase-frequency characteristic may be defined as the dependence of the acoustic path on the frequency l(ω). This dependence may be caused by the curving the sound wave front due to diffraction on the user's head.

(175) To sum up, the analysis results show the following:

(176) (a) the microphones 34, 35, 36 provide maximum SNR in the frequency range below approximately 1.5 kHz and they are optimal for use as the main microphones;

(177) (b) the microphones 37, 38 provide minimal SNR in the frequency range over approximately 1.5 kHz and they are optimal for use as the noise microphones;

(178) (c) the microphones 33 provide intermediate SNR in the frequency range over approximately 1.5 kHz and they are not optimal for use as the noise microphones.

(179) This means that when speech information is received with no background noise, the microphones 34-36 provide the best quality of the target signal. The microphones 37, 38 provide the target signal with substantially depressed level and shifted phase in the frequency range over approximately 1.5 kHz, comparatively to the microphones 34-36.

(180) When diffused noise is received from a distant source and/or when the noise is received in a room where reflections occur, the levels of the noise signal obtained from each of the microphones are approximately equal. In other words, by placing the noise microphone on the back part of the user's neck, it is possible to obtain a signal having minimal content of the diffracted speech sound and about the same noise content as the main microphone. This facilitates further processing the signal and separating the noise content in order to provide desired SNR of the target signal.

(181) FIG. 15 and FIG. 25 shows a wearable telecommunication device, comprising a neck-wearable housing 1 with an electronic unit 9 attached thereto, two in-ear earphones, two cords 4A, 4B, one of which connects to one of the in-ear earphone to the electronic unit, and the other cord connects the other in-ear earphone to the electronic unit; and a microphone array for picking up and processing a user's voice, the microphone array comprising a front microphone 25, a rear microphone 37 and a processor, wherein the two cords are mechanically connected to the neck-wearable housing, and points of connection of the two cords to the neck-wearable housing are close to each other and form a dorsal cord connection node 5, and are further mechanically connected to each other in sections between the in-ear earphones and the dorsal cord connection node 6 to form a connection point, wherein the suboccipital cord connection node, the dorsal cord connection node, and an area of the housing close to the dorsal cord connection node form a rear portion of the wearable telecommunication device, wherein the rear microphone 37 is fixed on the rear portion; and wherein a front-facing portion of the neck-wearable housing is in contact with an upper chest when worn by the user.

(182) When the rear microphone 37 is a part of a microphone array, the rear microphone signal may be used in processing a front microphone signal. In particular, the noise signal may be estimated and the estimation may be subtracted from the front microphone signal by means of direct subtraction or spectrum-dependent subtraction as shown in FIG. 24A and FIG. 24B. Further the target signal may be processed by an active noise reduction (ANR) system which mathematically determines residual noise estimation and subtracts it from the target signal in order to additionally increase the SNR of the target signal.

(183) Generally, most of optimal filters applicable for the above-indicated task may be considered as instances of the Wiener filter (FIG. 26A). Two signals, x.sub.k and y.sub.k are simultaneously supplied to the filter inputs. The y.sub.k signal usually contains two signal components, one component correlating with the x.sub.k signal, and another component non-correlating with the x.sub.k signal. The Wiener filter allows optimal estimation of the y.sub.k signal component correlating with the x.sub.k signal, and forming an error signal by subtracting the estimation from the y.sub.k signal.

(184) An optimal finite impulse response (FIR) filter able to minimize a root-mean-square error may be defined based on the autocorrelation function of a reference signal and the cross-correlation function of the reference signal and the processed signal. Practically, the autocorrelation and cross-correlation functions are defined based on the previous samples of the corresponding signals, so a large amount of data is required for precisely calculating the estimation value.

(185) Another approach for defining the filter tap weight factors is using adaptive algorithms. Instead of forming a data set for defining the correlation functions and using these functions for calculating each single tap weight vector for approximation of the optimal Wiener filter, the data is sequentially used for adjusting the filter tap weights in the direction of the gradient minimizing the root-mean-square error. Generally, the filter tap weights are adjusted in response for every data set, so this method requires substantially less adaptation calculations for each sample, in comparison with optimal algorithms. Like an optimal filter in case of a stationary signal, an adaptive filter automatically adjusts the filter tap weights upon drifting correlation functions of the input signals. An adaptive filter is able to track statistical parameters of non-stationary signals, if the parameters change rate is slower than the convergence rate of the adaptive filter.

(186) The most commonly used adaptation algorithm for FIR filters uses a quadratic error surface for such filters. When the filter tap weights change by a low value, being inversely proportional to the local gradient of the filter tap weight objective function, then the tap weights tend to shift to the global minimum position of the error surface.

(187) The Widrow-Hoff Algorithm has proposed tap weight adaptation method for each sample, by using an instant gradient assumption (this method sometimes is referred to as a stochastic gradient method), instead of slow filter adjustment using an average gradient assumption, which may also be used in the invention.

(188) The adaptive algorithm may be defined as follows:
w(n+1)=w(n)+μx(n)e(n)  (41),

(189) where w is the adaptive filter tap weight, n is the step number, μ is the filter convergence factor defining stability and the convergence rate of the filter. This algorithm is knows as the least-mean-square (LMS) algorithm. The LMS algorithm is simple, numerically stable and it is widely used in various adaptive systems. The diagram in FIG. 26B illustrates possible embodiment of the LMS algorithm in a filter, where the product of multiplication of the error signal and the reference signal is used for adaptation of the filter tap weights.

(190) The main advantage of the LMS algorithm is its ultimate calculation simplicity, as just N+1 pairs of multiplication-addition actions shall be performed in each step for adjusting the filter tap weights. The reverse side of the simplicity is slow convergence and comparatively high error dispersion in the steady state, as the filter tap weights always fluctuate around optimal values, thus increasing the output noise.

(191) Applicability of a certain algorithm depends on numerous factors and shall be defined individually in each case, based on experimental data. For instance, when an adaptive algorithm like the LMS algorithm is used, extraction of the input signal components being correlated and non-correlated to each other is performed in the real-time mode. It is obvious that when the above-stated microphone layout comprising a front microphone and a rear microphone is used, all components in the frequency range below 1.7 kHz will be suppressed. The non-correlation degree of the front and rear microphone signals increases almost proportionally as the frequency rises in the range above 1.7 kHz, so the efficiency of the target signal detection increases correspondingly.

(192) There are numerous LMS algorithm modifications aimed at increase of the convergence rate or at decrease of the number of necessary calculations. The convergence rate increase may be attained by improvement of the gradient assumption as well as by transformation of the input signal so as to make its samples non-correlated. Decreasing the calculation complexity may be achieved, e.g., by using signs of the error signal and the time-delay line content instead of their values. This approach allows getting rid of multiplication operations while updating the filter tap weights.

(193) Loss of the target signal in the frequency range below 1.7 kHz may be compensated by digital filtration methods. However, the SNR of harmonic voice components will be worse in this case.

(194) Further improvement of the noise reduction system is introducing a third front microphone placed in a neck-worn device and located on the user's chest. This approach allows forming a full-size microphone array which provides solutions for a number of problems, e.g., dynamic forming the array directional diagram depending on the user's head rotation and increasing quality of the speech signal owing to widening the array aperture.

(195) FIG. 27 and FIG. 28 show an embodiment of the invention in a form of a wearable telecommunication device, comprising a neck-wearable housing 1 configured to be mounted on a human body and in contact with back, left, right sides of the neck and upper chest and having at least one electronic unit 9 attached thereto, two in-ear earphones 3A, 3B, two cords 4A, 4B, one of which connects to one of the in-ear earphone to the electronic unit, and the other cord connects the other in-ear earphone to the electronic unit, a microphone array for picking up and processing a user's voice, comprising a front microphone 34, a rear microphone 38 and processor, where the two cords are mechanically connected to the neck-wearable housing and points of connection of the two cords to the neck-wearable housing are close to each other and form a dorsal cord connection node 5, and are further mechanically connected to each other in sections between the in-ear earphones and the dorsal cord connection node to form a suboccipital cord connection node 6 at the connection point, where the rear microphone 34 is on a portion of the neck-wearable housing configured to be in contact with a back of the neck when worn by the user; and where a correlated portion of the signal from the rear microphone and the signal from the front microphones represents noise, while an uncorrelated portion of the signal represents useful data.

(196) The rear microphone may be fixed on the suboccipital cord connection node (37 on FIGS. 28, 29A), on the neck-wearable housing close to the dorsal cord connection node (38 on FIG. 29A), and between the suboccipital cord connection node and the dorsal cord connection node. The front microphone 25 is fixed on a portion of the neck-wearable housing that is in contact with the user's chest when worn by the user (FIG. 15).

(197) In some embodiments (FIG. 29B) the wearable device comprises two front microphones 36A, 36B wherein the two front microphones are fixed on a portion of the neck-wearable housing that is in contact with the user's chest and at a substantially the same height when worn by the user. In particular, the front microphone 36B is positioned on the left side of the user's chest, the front microphone 36A is positioned on the right side of the user's chest. The rear microphone 38 is positioned in close proximity to the dorsal node (FIGS. 28, 29A, 29B).

(198) This configuration allows implementation of an adaptive algorithm defining the correlated and non-correlated components of the signal separately for two pairs of microphones, front right—rear and front left—rear, which facilitates substantially increasing precision of the noise signal assumption and improving SNR of the target speech signal.

(199) The device may further include a phased or a gradient microphone array comprising at least two microphones fixed on a chest portion of the neck-wearable housing, the additional microphone array may be used to determine a directional diagram of received sound waves.

(200) Additionally, the target speech signal SNR may be improved by introducing one more microphone disposed below the pair of the front-right and front-left microphones so as to form a triangle (like shown in FIG. 22B), or by introducing two more microphones 34, 35 disposed below the pair of front-right and front-left microphones so as to form a tetragon (like shown in FIG. 29B). This configuration provides further improvement of the signal-to-noise ratio owing to an increase in the number of microphones in the array and possibly using more complicated adaptation algorithms, e.g., a phased sub-array or a gradient sub-array may be formed so as to determine the direction of the target sound and the direction of the noise sound.

(201) FIG. 29A and FIG. 29B show an embodiment of the invention in a form of a wearable electronic device comprising a neck-wearable housing 1 with at least one electronic unit 9 attached thereto, two in-ear earphones 3A, 3B, two cords 4, and a microphone array comprising four front microphones 34, 35, 36A, 36B and two rear microphones 37, 38, where the front microphones 36A, 36B are positioned on a substantially horizontal line (the first line), the front microphones 34, 35 are positioned on another substantially horizontal line below the first line. The neck-wearable housing is configured so as the front microphones 34, 35 are disposed in close vicinity of the left and right clavicles, correspondingly.

(202) The dorsal node 5 and the suboccipital node 6 comprise the two rear microphones 38, 37, correspondingly. The microphone array forms a number of signals further used for determination of the correlated and non-correlated components of the front and rear microphone signals, while the correlated components are considered as noise. The adaptive algorithms used for determination of the correlated and non-correlated components may be narrow-band algorithms or wide-band algorithms. The narrow-band adaptive algorithms are far simpler in implementation than the wide-band adaptive algorithms. Whether an algorithm shall be implemented as narrow-band or wide-band is determined based on the phase incursion between the microphones in a certain frequency range, see [8]:

(203) $\begin{array}{cc}\mathrm{\Delta \phi }=2\pi \frac{\Delta \mathrm{Fs}.Math.L}{c}<<1,& \left(42\right)\end{array}$

(204) where ΔFs is the frequency range, L is the distance between the outermost microphones, c is the velocity of sound in the atmosphere (approximately 340 meters per second).

(205) In the device under consideration, ΔFs=Fmax−Fmin=3.1 kHz, where Fmax=3.4 kHz is the highest signal frequency, Fmin=300 Hz is the lowest signal frequency, L=0,2 M is the distance between the outermost microphones, so the phase incursion Δφ≈11.5 and the condition (42) is not met. Therefore, wide-band adaptive algorithms shall be used in the device under consideration.

(206) When the noise signal is adaptively suppressed, the target signal is somewhat attenuated as well, because the noise signal mainly differs from the target signal by the direction of incoming sound wave. If the direction of the target sound is known, then an adaptive Frost algorithm may be used, see [6], which limits attenuation of the target signal. When a front microphone is positioned on the user's head (like a boom microphone), the adaptive Frost algorithm may be used fairly easily. However, when one or more front microphones are placed in the device and positioned on the user's chest, the direction of incoming target sound may be determined very roughly within a rather wide angle range due to possible rotation of the user's head. Nevertheless, the adaptive Frost algorithm may be used if one of the front microphones is selected as the best one, based on the target signal volume.

(207) Another approach is using a processed composite signal formed from several front microphones as the front signal in the adaptive Frost algorithm. One other approach is using one or more front microphones disposed in additional devices like glasses, watches, bracelets, rings, etc., where the front microphones are wirelessly connected to the wearable device.

(208) FIG. 30 shows an embodiment of the invention, where two front microphones 42, 43 are located in the glasses frame 44 of an additional accessory, glasses 45. The microphones placed in the glasses are directed towards the user's mouth.

(209) Still another approach to adaptive filter improvement is detecting silence periods in the user's speech and calculating the filter tap weights during these gaps. However, this algorithm requires a silence sensor to detect the speech gaps. An accelerometer may be used as the silence sensor, when the accelerometer is located in a portion of the device that is adjacent to the user's body surface, e.g., in the area of the temporomandibular joint. Alternatively, the earphones may comprise the silence sensor adjacent to the user's external auditory canal. Yet this method has some limitations diminishing its advantages. First, any non-intentional movement of the sensor relative to the user's body (like hiccups, yawning, chewing, etc.) results in rustles that are detected as the user's speech, and loose engagement of the sensor (e.g., due to bristle, skin roughness, skin and intra-ear secretions, etc.) causes considerable malfunction of such noise reduction systems.

(210) Using at least two microphones, where some of them are front microphones and some are rear microphones, allows detecting the speech gaps by analyzing the signal level. When the signal level in all microphones of the array is similar (i.e., the signals are correlated), then a speech gap is detected and the filter tap weights may be adjusted, and when the signal level of the front microphone is substantially greater than the signal level of the rear microphone, then the target signal is detected and processed.

(211) Calculation of the filter tap weights may be performed during the speech gaps, based on a projection algorithm, see [7] having high efficiency, fast convergence and relatively low calculation complexity. Further the projection algorithm is briefly discussed.

(212) Each microphone signal spectrum can be determined based on the Fast Fourier Transform (FFT):

(213) 0$\begin{array}{cc}\mathrm{Sp}\left(\mathrm{mf},\mathrm{mi},k\right)=\underset{\mathrm{nt}=0}{\overset{\mathrm{NFTT}-1}{.Math.}}S\left(\left(\mathrm{nt}+\mathrm{NFTT}.Math.k\right).Math.\Delta T,\mathrm{mi}\right)e^{-i\frac{2\pi }{\mathrm{NFFT}}\mathrm{nt}.Math.\mathrm{mi}},& \left(43\right)\\ \mathrm{mf}\in \left[0,\mathrm{NFFT}/2\right],\mathrm{mi}\in \left[0,\mathrm{Mic}-1\right],k\in 0,1,2,\mathrm{.Math.},& \left(44\right)\end{array}$

(214) where Sp (mf, mi, k) is the signal spectrum value for the sample number mf for the microphone number mi and for the processing cycle number k; S ((nt+NFTT×k)×ΔT, mi) is the signal value for the time count number (nt+NFTT×k) and for the microphone number mi; ΔT=(Fd).sup.−1 is the sampling period in the analog-to-digital converter (ADC); Fd>2×Fmax is the sampling frequency in the ADC (Fmax is the highest target signal frequency); NFFT is the number of time counts while the signal spectrum is formed; Mic is the number of microphones in the device.

(215) Further the signal correlation matrix is formed for NumD frequency ranges:
K(mi,mi1,nd,k)=Σ.sub.kf=0.sup.MaxFD-1[Sp(kf+nd.Math.MaxFD,mi,k).Math.Sp(mf+nf.Math.MaxFD,mi1,k)*],ndε[0,numD−1],  (45),

(216) where MaxFD is the number of spectral samples in a frequency sub-range; NumD=round (NFFT×(Fmax−Fmin)/(Fd×MaxFD)) is the number of frequency sub-ranges in the operational bandwidth.

(217) The MaxFD value is determined so as to meet the following two conditions: the narrow-band condition
2πMaxFd×Fd×L/(NFFT×c)<<1  (46),

(218) and the stable estimation of the correlation matrix condition
MaxFD>2×Mic  (47).

(219) A weight factor is formed for every frequency sub-range so as to suppress the interfering noise signals:
{right arrow over (C)}(nd)=(I−K.sub.I(nd).Math.(K.sub.I(nd).Math.K.sub.I(nd)).sup.−1.Math.K.sub.I(nd).Math.{right arrow over (S)}(nd)  (48),

(220) where I is a unitary matrix having dimension Mic; K.sub.I(nd) is the matrix composed of I columns of the correlation matrix (45); {right arrow over (S)}(nd) is the reference vector having dimension Mic and providing directing the microphones to the user's mouth in the frequency sub-range nd.

(221) The vector element for the microphone number mi is:

(222) $\begin{array}{cc}S\left(\mathrm{nd},\mathrm{mi}\right)=K_{\mathrm{mi}}{e}^{-i\frac{2\pi .Math.R_{\mathrm{mi}}}{c}.Math.\left(\mathrm{Fmin}+\frac{\mathrm{Fd}.Math.\mathrm{MaxFD}}{\mathrm{NFFT}}\mathrm{nd}\right)},& \left(49\right)\end{array}$

(223) where Rmi is the distance between the user's mouth and the microphone number mi; Kmi is sensitivity of that microphone depending on the microphone position.

(224) The above relatively simple expression is generally not valid when diffraction occurs, as the wave front is substantially corrupted. Therefore, the above model is satisfactorily describes the signals obtained from microphones located in the area of ideal, non-distorted wave front as defined in the expression (40). However, according to experimental data, the expression (49) still may be used for low frequency signals even in diffracted area, as the sound phase is almost linearly depends on frequency (FIG. 23B).

(225) The resulted signal spectrum is formed as follows:

(226) $\begin{array}{cc}\mathrm{Sp}0\left(\mathrm{kf}+\mathrm{Max}\mathrm{FD}.Math.\mathrm{nd},k\right)=\underset{\mathrm{mi}=0}{\overset{\mathrm{Mic}-1}{.Math.}}{C}^{*}\left(\mathrm{mi}\right).Math.\mathrm{Sp}\left(\mathrm{kf}+\mathrm{Max}\mathrm{FD}.Math.\mathrm{nd},\mathrm{mi},k\right),\mathrm{kf}\in \left[0,\mathrm{Max}\mathrm{FD}\right],\mathrm{nd}\in \left[0,\mathrm{NumD}-1\right],& \left(50\right)\end{array}$

(227) The resulted time-domain signal is formed, based on the Inverse Fourier Transform (IFT):

(228) $\begin{array}{cc}S0\left(\mathrm{nt},k\right)=\underset{\mathrm{mf}=0}{\overset{\mathrm{NFFT}/2}{.Math.}}\mathrm{Sp}\left(\mathrm{mf},k\right)e^{i\frac{2\pi }{\mathrm{NFFT}}\mathrm{nt}.Math.\mathrm{mf}},\mathrm{nt}\in \left[0,\mathrm{NFFT}-1\right],k\in 0,1,2,\mathrm{.Math.}.& \left(51\right)\end{array}$

(229) When the process is divided into two stages as described in the above, the subtraction of the noise signal may advantageously be provided on the first stage, when the user speaks, whereas selection of the weight factors and direction vectors (see the expression (49)) may be performed on the second stage during the speech gaps, which allows substantial improving the noise reduction efficiency.

(230) Obviously, if a microphone of the microphone array shall be positioned as a rear microphone, it is expedient to place it in a neck-worn device. The neck-worn device may be implemented as a wireless headset, a wearable electronic device, a wearable multimedia device, a wearable personal computer, a hearing aid, etc. provided in a form of a necklace or otherwise comprising a neck-wearable housing, wherein the front microphones are positioned on the user's chest, while the rear microphone or microphones may be positioned on the back surface of a helmet, a headwear piece, a shirt collar, or in a rear cord connection node of the device, which is preferable embodiment of the invention.

(231) The neck-worn device may comprise a neck-wearable housing in any form. The neck-worn device may be based on an O-shaped loop as shown in FIG. 29B, or it may be based on a U-shaped loop (FIG. 15) comprising the same front microphones.

(232) In this discussion, neck-wearable housing of the claimed device is mainly described as O-shaped or U-shaped loop in examples and embodiments. However, other types of neck-wearable housing may be possible.

(233) In many embodiments the neck-wearable housing may be flexible in at least one location, for example the portions of a neck-wearable housing between dorsal node 5 and electronic unit 9 are flexible (FIGS. 15, 28). Also, the entire housing may be flexible as well.

(234) As the position of the rear microphone is predetermined and fixed in relation to the front microphones of the wearable device, and the distance between the rear microphone and the target sound source is known as well (at least roughly), then diffraction effects occurring while sound waves pass around a human head may be considered as accountable factors, see [11].

(235) However, positioning the rear microphone under clothes causes substantial distortion of the sound received by this microphone. This is why positioning the rear microphone in a rear cord connection node is ideal in view of providing the best possible noise reduction processing of the target signal. For example, when the rear microphone 37 is placed in the suboccipital node 6 (FIG. 25), its position is fixed in relation to the front microphones, no matter if the user's head is rotated or not, as the earphone cords adhere to the user's head by tension force and the suboccipital node moves along with the corresponding point of the back surface of the user's neck. Moreover, in this case, the rear microphone is not covered by clothes even if the neck loop itself is located under the clothes.

(236) Alternatively, the rear microphone 38 may be placed in the dorsal node 5 (FIG. 29A). This option may be advantageous when the user has long hair or uses a headwear article (typically like female users) which may cover the suboccipitally positioned microphone and impede penetration of acoustic waves. Moreover, this option may be still advantageous even when the user has short hair or a bald head, as the rear microphone placed in the dorsal node is better protected against wind by the clothes.

(237) In one embodiment of the invention, the rear microphone 38 may be movably positioned between the suboccipital 6 and dorsal 5 nodes (FIG. 31), so the user is able to choose and use the best rear microphone position, depending on the operation conditions (the type of used clothes, wind, rain, acoustic environment, etc.). This option requires a separate cord for the rear microphone (like the helical cord 46 in FIG. 31), but such a feature may hardly be considered as a substantial drawback in view of the obvious advantages thereof.

(238) In another embodiment of the invention, the rear microphone may be selected among the two rear microphones, where one of them is located in the suboccipital node and another one is located in the dorsal node, depending on the device operation conditions. For instance, when the user walks in the street, the suboccipital node position may be preferable for the rear microphone, and when the user drives a car, the dorsal node position may be preferable due to proximity of the head restraint. Sometimes, the earphones may be retracted, so the dorsal node position and the suboccipital node position may be positioned fairly close to each other. This option is illustrated by FIG. 32 and FIG. 33. The selection may be done manually by the user (e.g., via the device user interface) or automatically, based on the analysis and comparison of quality of signals of the rear microphones.

(239) In still another embodiment of the invention, the rear microphone signal may be composed of the signals obtained from two rear microphones, e.g., using one or more of the processing algorithms described in the above.

(240) Thus, owing to the necklace-like form-factor of the wearable device and the presence of the suboccipital and dorsal cord connection nodes, the front microphones may be positioned in the area of direct propagation of the speech sound wave, while the rear microphone (or several rear microphones) used as the noise sound wave receiver(s) may be positioned on the back surface of the user's neck in the area of the acoustic shadow for the speech sound wave. This layout pattern allows forming a correlation microphone array, wherein components of the front signal correlated with the rear signal may be treated as noise, while components of the front signal non-correlated with the rear signal may be treated as the target signal.

(241) The first step of the signal processing includes forming a composite front signal comprising an additive mix of a noise signal and a target signal, obtained from the front microphone(s). The rear microphone(s) form(s) a rear signal comprising mainly a noise signal. Further, both signals are processed by an adaptive digital filter (e.g., a Wiener-like filter) so as to extract the target signal, using the LMS method. The system may be described by the following equations:
e.sub.n=d.sub.n−y.sub.n  (52),
y.sub.n=w.sub.n.sup.Tx.sub.n  (53)

(242) where e.sub.n is the filtered target signal; d.sub.n is the combined front signal; y.sub.n is the filtered noise signal; w.sub.n.sup.T is a transposed matrix of adaptive filter tap weights w.sub.n; x.sub.n is the combined rear signal.

(243) The adaptation principle is described as follows:
w.sub.n+1=w.sub.n+μe.sub.nx.sub.n  (54),

(244) where n is the adaptation step number; μ is a positive constant defining stability and the convergence rate of the algorithm.

(245) The correlation between the input signal x.sub.n and the output signal d.sub.n, may be presented as a discrete transfer function. After successful adaptation, W(z) acceptably approximates the transfer function, so the adaptive filter may identify the system transfer function.

(246) In the Filtered-x LMS algorithm (e.g., see [12]), P(z) is the system transfer function, {circumflex over (P)}(z) is the system transfer function model obtained by the identification. In this case the system is described as follows:
e.sub.n=d.sub.n−P(z)y.sub.n  (55),

(247) where y.sub.n is the same as in the expression (53), but the expression (54) is modified in the following way:
w.sub.n+1=w.sub.n+μe.sub.nr.sub.n  (56),

(248) where r.sub.n is a vector formed from current and previous samples of the filtered reference signal r.sub.n
r.sub.n={circumflex over (P)}(z)x.sub.n  (57).

(249) The function {circumflex over (P)}(z) may be presented either by a finite-impulse response (FIR) filter or an infinite-impulse response (IIR) filter; however, FIR filters are used more commonly owing to their higher stability. Identification of the function P(z) may be performed by a usual LMS algorithm. When the identifying signal is a white noise, then W(z) acceptably accurate approximates P(z). The system is considered stable if the phase error of the model does not exceed π/2. Thus, the output signal may be formed as defined in the expression (52).

(250) In real operating conditions, the disturbance and the system response cannot be measured with ideal accuracy, so iterative methods can be used in order to maintain the algorithm convergence. This approach may be advantageous when processing non-stationary signals, though the signal duration shall exceed the disturbance duration.

(251) Having thus described a preferred embodiment, it should be apparent to those skilled in the art that certain advantages of the described method and apparatus have been achieved.

(252) It should also be appreciated that various modifications, adaptations, and alternative embodiments thereof may be made within the scope and spirit of the present invention. The invention is further defined by the following claims.

LIST OF CITED REFERENCES

(253) 1. J. Benesty, J. Chen, Y. Huang, I. Cohen, “Noise reduction in speech processing”, Springer Science & Business Media (Apr. 28, 2009). 2. O. Hirsch, “Bluetooth microphone array”, US 2010/0048131 A1 (2010). 3. G. C. Burnett, “Microphone array with rear venting”, US 2009/0003640 A1 (2009). 4. A. L. Ushakov, “Headset for a mobile electronics”, US 2014/0185821 A1 (2014). 5. Hyun Jun Park, Kwokleung Chan, “Systems, methods, apparatus and computer program products for enhanced active noise cancellation” US 2010/0131269 A1 (2010) 6. J. Benesty, J. Chen, Y. Huang “Microphone Array Signal Processing”, Springer Topics in Signal Processing (2008). 7. Malyshkin, G. S., Optimal and adaptive methods of processing of hydroacoustic signals, v. 1, v. 2, St. Petersburg, “Electropribor” publishing, 2009. 8. Monsingo, R. A., Miller, T. U. Adaptive antenna arrays. Introduction to Theory, Radio and Communications, 1986, p. 448. 9. Sementsov et al., Research into acoustic transfer functions and speech clarity for wearable devices, www.necktec.ru, downloaded 22 Nov. 2015 10. Irina Aldoshina, Bi-aural Current status of research, http://www.show-master.ru/ Downloaded 22 Nov. 2015 11. Mayer, R. Numerical methods of solving of edge cases or a vibrating membrane, http://mathhelpplanet.com/ Downloaded 22 Nov. 2015 12. Asutosh Kar, A bika Prasad Chanda Sarthak Mohapatra Mahesh Chandra An Improved Filtered-x Least-mean-square Algorithm for Acoustic Noise Suppression.//Advanced Computing and Informatics. Proceedings of the Second International Conference on Advanced Computing, Networking and Informatics (ICACNI-2014), vol. 1 pages 25-32, Springer International Publishing (2014).