Speech enhancement method and apparatus for same

Abstract

A speech enhancement method is provided. The speech enhancement method includes: estimating a direction of a speaker by using an input signal, generating direction information indicating the estimated direction, detecting speech of a speaker based on a result of the estimating the direction, and enhancing the speech of the speaker by using the direction information based on a result of the detecting the speech.

Claims

1. An electronic device comprising: a plurality of microphones disposed on an upper part of the electronic device; a processor configured to: receive at least one signal via any one or any combination of the plurality of microphones; estimate a direction of the received at least one signal; determine whether the estimated direction of the received at least one signal is a predetermined direction; based on the estimated direction of the received at least one signal being determined to be the predetermined direction, determine whether the received at least one signal is associated with a speech of a user to control an operation of the electronic device; based on the received at least one signal being determined to be associated with the speech of the user to control the operation of the electronic device, beamform the received at least one signal such that the received at least one signal from the predetermined direction is emphasized; output beamformed at least one signal to perform speech recognition; and based on the estimated direction of the received at least one signal being determined to not be the predetermined direction, deactivate the determining whether the received at least one signal is associated with the speech of the user, the beamforming of the received at least one signal, and the outputting of the received at least one signal.

2. The electronic device of claim 1, wherein the electronic device further comprises a screen, wherein the processor is further configured to determine whether the estimated direction of the received at least one signal is the predetermined direction, based on a direction of the speech when the user speaks while viewing the screen of the electronic device.

3. A speech enhancement method being performed by an electronic device, the speech enhancement method comprising: receiving at least one signal via any one or any combination of a plurality of microphones; estimating a direction of the received at least one signal; determining whether the estimated direction of the received at least one signal is a predetermined direction; based on the estimated direction of the received at least one signal being determined to be the predetermined direction, determining whether the received at least one signal is associated with a speech of a user to control an operation of the electronic device; based on the received at least one signal being determined to be associated with the speech of the user to control the operation of the electronic device, beamforming the received at least one signal such that the received at least one signal from the predetermined direction is emphasized; outputting beamformed at least one signal to perform speech recognition; and based on that the estimated direction of the received at least one signal being determined to not be the predetermined direction, deactivating the determining whether the received at least one signal is associated with the speech of the user, the beamforming of the received at least one signal, and the outputting of the received at least one signal.

4. The electronic device of claim 3, wherein the determining whether the estimated direction of the received at least one signal is the predetermined direction is based on a direction of the speech when the user speaks while viewing a screen of the electronic device.

5. A non-transitory computer-readable storage medium storing instructions that cause a processor of an electronic device to: receive at least one signal via any one or any combination of a plurality of microphones disposed on an upper part of the electronic device; estimate a direction of the received least one signal; determine whether the estimated direction of the received at least one signal is a predetermined direction; based on the estimated direction of the received at least one signal being determined to be the predetermined direction, determine whether the received at least one signal is associated with a speech of a user to control an operation of the electronic device; based on the received at least one signal being determined to be associated with the speech of the user to control the operation of the electronic device, beamform the received at least one signal such that the received at least one signal from the predetermined direction is emphasized; output the beamformed at least one signal to perform speech recognition; and based on that the estimated direction of the received at least one signal being determined to not be the predetermined direction, deactivating the determining whether the received at least one signal is associated with the speech of the user, the beamforming of the received at least one signal, and the outputting of the received at least one signal.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

(1) FIG. 1 is a schematic block diagram of a speech enhancement apparatus according to an exemplary embodiment;

(2) FIG. 2 is a schematic block diagram of a speech enhancement apparatus according to another exemplary embodiment;

(3) FIG. 3 is a flowchart of a speech enhancement method according to an exemplary embodiment;

(4) FIG. 4 is a flowchart of a speech enhancement method according to another exemplary embodiment;

(5) FIG. 5 illustrates a speech enhancement apparatus for mobile phone, according to an exemplary embodiment;

(6) FIG. 6 illustrates a speech enhancement apparatus for a TV, according to an exemplary embodiment;

(7) FIG. 7 illustrates a speech enhancement apparatus for a tablet, according to an exemplary embodiment;

(8) FIG. 8 illustrates a speech enhancement apparatus for a wearable device, according to another exemplary embodiment; and

(9) FIG. 9 illustrates a speech enhancement apparatus for a headset or hearing aid, according to an exemplary embodiment.

DETAILED DESCRIPTION OF THE EMBODIMENTS

(10) The advantages and features of the inventive concept and methods of achieving the advantages and features will be described more fully with reference to the accompanying drawings, in which exemplary embodiments of the inventive concept are shown. The inventive concept may, however, be embodied in many different forms and should not be construed as being limited to the exemplary embodiments set forth herein; rather these exemplary embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the concept of the inventive concept to one of ordinary skill in the art.

(11) Hereinafter, the terms used in the specification will be briefly described, and then the inventive concept will be described in detail.

(12) The terms used in this specification are those general terms currently widely used in the art in consideration of functions in regard to the inventive concept, but the terms may vary according to the intention of those of ordinary skill in the art, precedents, or new technology in the art. Moreover, specified terms may be selected by the applicant, and in this case, the detailed meaning thereof will be described in the detailed description of the inventive concept. Thus, the terms used in the specification should be understood not as simple names but based on the meaning of the terms and the overall description of the inventive concept.

(13) Throughout the specification, it will also be understood that when a component “includes” an element, unless there is another opposite description thereto, it should be understood that the component does not exclude another element but may further include another element. In addition, terms such as a “ . . . unit” used in the specification refers to a hardware component such as software or hardware component such as FPGA or ASIC, and a “ . . . unit” performs certain functions. However, a “ . . . unit” is not limited to software or hardware. A “ . . . unit” may be configured in an addressable storage medium or to reproduce one or more processors. Thus, for example, a “ . . . unit” includes components such as software components, object-oriented software components, class components, task components, processes, functions, attributes, procedures, subroutines, segments of programs codes, drivers, firmware, microcodes, circuits, data, databases, data structures, tables, arrays, and variables. Functions provided in the components and the “ . . . units” may be combined with a smaller number of components and “ . . . units” or separated from additional components and “ . . . units.”

(14) Hereinafter, the inventive concept will now be described more fully with reference to the accompanying drawings, in which exemplary embodiments are shown such that one of ordinary skill in the art may easily understand the inventive concept. Details that are not related to description of the inventive concept will be omitted for clarity of description.

(15) A user wishes to use a device regardless of where and in what posture the device is. Exemplary embodiments of the inventive concept are needed to apply a speech enhancement technique using microphones, to actual devices. Exemplary embodiments of the inventive concept relate to the enhancement of speech by searching for a direction of a user at the moment when the user speaks.

(16) FIG. 1 is a block diagram of a speech enhancement apparatus 10 according to an exemplary embodiment.

(17) Referring to FIG. 1, the speech enhancement apparatus 10 includes a sensor unit 11 (e.g., a sensor, etc.), a speech detection unit 12 (e.g., a speech detector, etc.), a direction estimation unit 13 (e.g., a direction estimator, etc.), and a speech enhancement unit 14 (e.g., a speech enhancer, etc.).

(18) The sensor unit 11 includes at least one microphone. The sensor unit 11 may further include a camera, a bone conduction sensor, a proximity sensor, an infrared sensor, an acceleration sensor, or an ultrasonic sensor in addition to the at least one microphone.

(19) The speech detection (SD) unit 12 detects speech of a speaker by using a signal of the sensor unit 11. For example, the speech detection unit 12 detects whether a user has spoken by using a microphone or other sensors.

(20) The direction estimation unit 13 (target direction estimation (TDE) estimates a direction of a speaker by using a signal of the sensor unit 11 and generates direction information indicating the estimated direction. For example, the direction estimation unit 13 detects a position of a user by using multiple microphones or an image of a camera.

(21) The speech enhancement (ENH) unit 14 enhances speech of a speaker that is input to a microphone. In detail, the speech enhancement unit 14 receives information indicating whether the user has spoken and direction information, and enhances a speech signal by using the information indicating whether the user has spoken and the direction information.

(22) The speech detection unit 12 controls operations of the direction estimation unit 13 and the speech enhancement unit 14 based on a result of speech detection. Control information C1 transmitted from the speech detection unit 12 to the direction estimation unit 13 may be used to control switching the direction estimation unit 13 on or off. Control information C2 transmitted from the speech detection unit 12 to the speech enhancement unit 14 may control a filter update of the speech enhancement unit 14.

(23) The direction estimation unit 13 transmits direction information to the speech enhancement unit 14. Information T transmitted from the direction estimation unit 13 to the speech enhancement unit 14 denotes direction information of a user.

(24) As an input transmitted from the sensor unit 11 to the speech detection unit 12, a camera or bone conduction information may be used besides a microphone signal. As an input transmitted from the sensor unit 11 to the direction estimation unit 13, camera sensor information may be used in addition to a microphone signal. An input transmitted from the sensor unit 11 to the speech enhancement unit 14 may be a microphone signal.

(25) Examples of methods of detecting speech by using the speech detection unit 12 include detecting whether there is a speech component by using a microphone signal, detecting movement of the lips of a speaker by using a camera signal in addition to voice activity detection (VAD) in which speech is used, detecting whether a speaker has spoken by using a bone conduction signal, and detecting speech in a multimodal manner in which the above methods are combined.

(26) Examples of estimating a direction of speech by using the direction estimation unit 13 include estimating a direction by using a difference between two or more microphone signals, estimating a direction of arrival (DOA) or time difference of arrival (TDOA) using a speech, recognition of a face of a speaker, detecting movement of the lips of a speaker, recognition of gestures, or in a multimodal manner in which these methods are combined.

(27) The speech enhancement unit 14 may include at least one filter. When speech is detected by the speech detection unit 12, a filter factor is counted, and when no speech is detected by the speech detection unit 12, noise for noise modeling may be estimated. The speech enhancement unit 14 may use direction information received from the direction estimation unit 13 to perform at least one of adjustment of time when a desired signal source arrives at each microphone, correction of a deviation between microphones, and separation of signal sources.

(28) The speech enhancement apparatus 10 may operate also as follows. When a user speaks while a speech recognition mode or a call mode is prepared, the speech detection unit 12 detects the speech. In a section in which an utterance of the user is detected, the speech detection unit 12 turns on the direction estimation unit 13 by using the control information C.sub.1 to search for a direction of the user while the user is speaking, and transmits the direction information T to the speech enhancement unit 14. Here, filter update of the speech enhancement unit 14 may be blocked to prevent speech distortion. The speech enhancement unit 14 corrects a delay between channels by using the received direction information T, and performs speech enhancement by adjusting a filter update by using the control information C.sub.2.

(29) According to the speech enhancement apparatus 10, the direction estimation unit 13 searches for a direction of the user only when the speech detection unit 12 detects speech, and thus, an exact direction of the user may be determined. If a direction is searched for while no speech has been uttered, an inaccurate direction may be estimated due to noise or the like. The speech enhancement unit 14 may perform speech enhancement by using exact direction information. Thus, by using the speech enhancement apparatus 10, speech enhancement performance by using direction estimation performance and direction information may be improved. Moreover, if speech is not detected, operations of the direction estimation unit 13 and the speech enhancement unit 14 may be stopped, and thus, a gain may be obtained also for computing power of the speech enhancement apparatus 10.

(30) FIG. 2 is a schematic block diagram of a speech enhancement apparatus 20 according to another exemplary embodiment.

(31) Referring to FIG. 2, a sensor unit 21, a speech detection unit 22, a direction estimation unit 23, and a speech enhancement unit 24 of the speech enhancement apparatus 20 are similar to the sensor unit 11, the speech detection unit 12, the direction estimation unit 13, and the speech enhancement unit 14 of the speech enhancement apparatus 10 of FIG. 1, respectively. Hereinafter, description will focus on differences of the speech enhancement apparatus 20 from the speech enhancement apparatus 10.

(32) Structures of the speech detection unit 22 and the direction estimation unit 23 of the speech enhancement apparatus 20 are different from those of the speech detection unit 12 and the direction estimation unit 13 of the speech enhancement apparatus 10. In detail, directions of control information C.sub.1 are opposite in the speech enhancement apparatus 10 and the speech enhancement apparatus 20.

(33) The direction estimation unit 23 controls an operation of the speech detection unit 22 based on a result of direction estimation. Control information C.sub.1 transmitted from the direction estimation unit 23 to the speech detection unit 22 may be used to control switching the speech detection unit 22 on or off.

(34) According to the speech enhancement apparatus 20, a direction may be first estimated to selectively process speech related to an operation of the speech enhancement apparatus 20 among detected speeches. For example, not only speech of a speaker to be input to a wearable device but daily conversation of the speaker may also be continuously input to a wearable device which is in the form of a watch. Here, the daily conversion of the speaker is highly likely not to be related to an operation of the wearable device. Speech of a speaker related to an operation of the wearable device is highly likely to be speech of the speaker while viewing a screen of the wearable device. Thus, according to the speech enhancement apparatus 20, if the direction estimation unit 23 estimates a direction of a speaker, based on camera sensor information or the like, that is, if the speaker is viewing a screen of the wearable device, the speech detection unit 22 detects speech so as to select speech that is highly related to use of the speech enhancement apparatus 20. Further, if no direction is estimated, operations of the speech enhancement unit 23 and the speech enhancement unit 24 may be stopped, and thus, a gain may be obtained also for computing power of the speech enhancement apparatus 20.

(35) Examples of speech enhancement-related techniques may include beamforming whereby speech in a predetermined direction is focused and ambient noise is removed and noise suppression whereby stationary noise is removed.

(36) Hereinafter, exemplary embodiments of a speech enhancement method according to exemplary embodiments of the inventive concept will be described with reference to FIGS. 3 and 4.

(37) FIG. 3 is a flowchart of a speech enhancement method according to an exemplary embodiment.

(38) In operation 31, speech of a speaker is detected.

(39) In operation 32, whether speech is detected or not is determined. If it is determined that speech is not detected (NO), the method returns to operation 31.

(40) In operation 32, if it is determined that speech is detected (YES), a direction of a speaker is estimated by using an input signal in operation 33 and direction information T indicating the estimated direction is generated.

(41) In operation 32, if it is determined that speech is detected (YES), in operation 34, the speech of the speaker is enhanced by using the direction information T of operation 33.

(42) Thus, according to the speech enhancement method of FIG. 3, a direction is searched for only when speech is detected, and thus, a direction of the speaker may be determined. In addition, if it is determined that no speech is detected, direction estimation and speech enhancement may be omitted, and thus a gain may be obtained also for computing power.

(43) FIG. 4 is a flowchart of a speech enhancement method according to another exemplary embodiment.

(44) In operation 41, a direction of a speaker is estimated by using an input signal, and direction information indicating the estimated direction is generated.

(45) In operation 42, whether a direction has been estimated or not is determined. If it is determined that no direction is estimated (when operation 42 corresponds to no), the method returns to operation 41.

(46) In operation 42, if it is determined that a direction is estimated (when operation 42 corresponds to yes), speech of a speaker is detected in operation 43.

(47) In operation 44, whether speech is detected or not is determined. If it is determined that no speech is detected (when operation 44 corresponds to no), the method returns to operation 41.

(48) If it is determined that speech is detected (when operation 44 corresponds to yes), in operation 45, the speech of the speaker is enhanced by using direction information T of operation 41.

(49) Thus, according to the speech enhancement method of FIG. 4, speech is detected if a direction of a speaker is able to be estimated, and thus, speech, which is highly likely to be used in speech enhancement, may be selected. Moreover, if no direction is estimated, speech detection and speech enhancement may be omitted, and thus, a gain may be also obtained for computing power.

(50) Hereinafter, noise reduction in a speech enhancement apparatus according to an exemplary embodiment of the inventive concept will be described.

(51) Stationary noise of a predetermined level (for example, a sine wave and white noise) is reproduced under situations where directions of speakers and noise directions are different. For example, a level of noise is assumed to be 10.

(52) First, solution input and output are checked while there is no speech of a speaker. In this case, for example, an input signal noise level may be 10, and an output signal noise level may be about 8. Here, as speech of a speaker is not detected, direction estimation and speech enhancement are not performed; instead, noise modeling may be performed by using a speech enhancement unit.

(53) Next, solution input and output after a speaker has spoken are checked. In this case, for example, an input single noise level may be 10, and an output signal noise level may be about 3. Here, as speech of a speaker is detected, direction estimation and speech enhancement are performed. As speech enhancement is performed by using direction information generated based on direction estimation, the speech enhancement may be focused on a speaker direction to concentrate more on the speech of the speaker, and an influence of a signal in a noise direction may be reduced, and accordingly, a noise level in an output signal is reduced.

(54) Thus, according to the speech enhancement apparatus of an exemplary embodiment, even when noise of the same level is present and input before and after a speaker has spoken, there is a difference in output signal noise levels. When there is no speech of a speaker, noise of a predetermined level is output and thus noise modeling may be performed, and when there is speech of a speaker, a noise level is reduced and thus speech of a speaker may be enhanced.

(55) Hereinafter, various exemplary embodiments in which a speech enhancement apparatus of the inventive concept is implemented in a device will be described with reference to FIGS. 5 through 9.

(56) FIG. 5 illustrates a speech enhancement apparatus 10 for a mobile phone, according to an exemplary embodiment.

(57) The speech enhancement apparatus 10 of FIG. 5 is substantially the same as the speech enhancement apparatus 10 of FIG. 1 having a basic block structure. In a mobile phone of FIG. 5, microphones may be located at positions marked with a shaded circle as ∘. For example, microphones may be placed at an upper part and a lower part of the mobile phone. However, the positions and the number of microphones are not limited thereto, and microphones may be placed at other various positions. The speech detection unit 12 and the direction estimation unit 13 of the speech enhancement apparatus 10 for a mobile phone may use speech information of the microphones.

(58) FIG. 6 illustrates a speech enhancement apparatus 10 for a TV, according to an exemplary embodiment.

(59) In the example of a TV of FIG. 6, microphones may be located at positions marked with a shaded circle as ∘. For example, microphones may be placed at an upper edge of the TV. However, the positions and the number of microphones are not limited thereto, and microphones may be placed at other various positions. Further, the TV may include a camera that is not shown. The TV may perform a multimedia function such as a video call function by using a microphone and a camera. A user may recognize speech or make a video call in front of the TV. Here, the speech detection unit 12 and the direction estimation unit 13 of the speech enhancement apparatus 10 for a TV may use both speech information and image information of the camera.

(60) FIG. 7 illustrates a speech enhancement apparatus 10 for a tablet, according to an exemplary embodiment.

(61) In the example of a tablet of FIG. 7, microphones may be located at positions marked with a shaded circle as ∘. For example, microphones may be placed at a lower edge of a tablet. However, the positions and the number of microphones are not limited thereto, and microphones may be placed at other various positions. Usage behaviors of a tablet of a user vary, and thus a hybrid method according to usage behaviors may be selected. First, the speech detection unit 12 of the speech enhancement apparatus 10 for a tablet detects speech based on speech information. Next, if it is possible to perform face recognition based on camera information, like a scenario of a speech enhancement apparatus for TVs, the speech detection unit 12 and the direction estimation unit 13 of the speech enhancement apparatus 10 for a tablet may use both speech information and image information of a camera. If face recognition is not possible, like a scenario of a speech enhancement apparatus for mobile phones, the speech detection unit 12 and the direction estimation unit 13 of the speech enhancement apparatus 10 for a tablet may use speech information of the microphones.

(62) FIG. 8 illustrates a speech enhancement apparatus 20 for a wearable device according to another exemplary embodiment.

(63) In the case of a wearable device, when a user is using a wearable device, face of the user usually faces the wearable device. When the face of the user is facing the wearable device, a recognition rate of speech of the user may be increased. That is, in the case of a wearable device, a speech detection efficiency may be further increased by setting a condition that a preprocessing operation is performed only when a face is recognized by using camera information. Thus, the speech enhancement apparatus 20 for a wearable device of FIG. 8 has almost the same structure as that of the speech enhancement apparatus 20 of FIG. 2 having a block structure. In the example of a wearable device of FIG. 8, microphones may be located at positions marked with a shaded circle as ∘. For example, microphones may be placed at a left upper end and a right lower end of the wearable device. However, the positions and the number of microphones are not limited thereto, and microphones may be placed at other various positions.

(64) The direction estimation unit 23 of the speech enhancement apparatus 20 for a wearable device notifies whether a face is recognized or not by using camera image information. Next, if a face is recognized, the speech detection unit 22 and the speech enhancement unit 24 operate. If a face is not recognized, the speech detection unit 22 and the speech enhancement unit 24 do not operate. If the wearable device has a single microphone, direction information T of the direction estimation unit 23 is not transmitted to the speech enhancement unit 24.

(65) FIG. 9 illustrates a speech enhancement apparatus 10 for a headset or hearing aid, according to an exemplary embodiment.

(66) When a user speaks, the speech is transmitted through the air but also through, for example, the skull of the user. To a headset that is directly worn on the head of a user or a hearing aid that is directly worn in the ears of a user, bone conduction that is transmitted via the skull of the user may be detected. Thus, the speech detection unit 12 of the speech enhancement apparatus for a headset or hearing aid may detect speech by using bone conduction sensor information in addition to speech information.

(67) The inventive concept relates to a method of searching for a speaker direction at the moment when the user speaks to enhance a speech, wherein the method is a speech enhancement method that efficiently operates by distinguishing a speech section and a non-speech section and secures high performance. According to the inventive concept, the speech enhancement method may also be applied to TVs, tablets, or hearing aid besides mobile phones. In particular, body-attached type computing devices (wearable computer or wearable device) such as watches, rings, or glasses are gaining attention and appear in product forms in the industry, and if the method is combined with these devices, the method may contribute as an efficient low-power algorithm.

(68) The inventive concept can also be embodied as computer readable codes on a computer readable recording medium. The computer readable recording medium is any recording device that can store data which can be thereafter read by a computer system.

(69) Examples of the computer readable recording medium include read-only memory (ROM), random-access memory (RAM), CD-ROMs, magnetic tapes, floppy disks, optical data storage devices, etc. (such as data transmission through the Internet). The computer readable recording medium can also be distributed over network coupled computer systems so that the computer readable code is stored and executed in a distributed fashion. In addition, functional programs, codes, and code segments for accomplishing the inventive concept can be easily construed by programmers skilled in the art to which the inventive concept pertains.

(70) While this inventive concept has been particularly shown and described with reference to exemplary embodiments thereof, it will be understood by those of ordinary skill in the art that various changes in form and details may be made therein without departing from the spirit and scope of the inventive concept as defined by the appended claims. The exemplary embodiments should be considered in a descriptive sense only and not for purposes of limitation. Therefore, the scope of the inventive concept is defined not by the detailed description of the inventive concept but by the appended claims, and all differences within the scope will be construed as being included in the inventive concept.