Hearing apparatus with a facility for reducing a microphone noise and method for reducing microphone noise

Abstract

An input signal is provided with a low microphone noise in a hearing apparatus. The microphone noise in the input signal of the hearing apparatus is reduced, by the input signal being filtered by a Wiener filter, if a noise power determined at the input signal is smaller than a predetermined limit value. The Wiener filter is however deactivated, if the noise power is greater than the limit value or equal to the limit value.

Claims

1. A method for reducing inherent microphone noise generated independently of ambient noise in an input signal of a hearing apparatus, which comprises the steps of: filtering the input signal, received by a microphone of the hearing apparatus, via a Wiener filter if a noise power determined for the input signal is smaller than a predetermined limit value for assisting in reducing the inherent microphone noise; and deactivating the Wiener filter if the noise power is greater than the predetermined limit value or equal to the predetermined limit value for assisting in reducing the inherent microphone noise.

2. The method according to claim 1, which further comprises, for noise power-dependent deactivation, weighting an attenuation of the Wiener filter acting on the input signal with a weighting factor, which is a function of the noise power.

3. The method according claim 2, wherein the function forms a gradual transition between a completely active attenuation and a completely deactivated attenuation.

4. The method according to claim 1, which further comprises limiting the noise power to a predetermined highest value.

5. The method according to claim 1, which further comprises estimating the noise power for at least one signal part of the input signal on a basis of the signal part according to a statistical estimation method.

6. The method according to claim 1, which further comprises determining the noise power for at least one signal part of the input signal on a basis of a characteristic microphone noise curve.

7. The method according to claim 1, which further comprises defining the predetermined limit value on a basis of a characteristic curve of a microphone.

8. The method according to claim 1, which further comprises limiting attenuation of the Wiener filter acting on the input signal to a predetermined maximum attenuation value with an active Wiener filter.

9. The method according to claim 1, which further comprises back-scaling the noise power in dependence on a current value of a directional parameter.

10. The method according to claim 1, which further comprises limiting an attenuation of the Wiener filter acting on the input signal to a highest value in dependence on a current value of a directional parameter.

11. The method according to claim 1, wherein the predetermined limit value is dependent on a current value of a directional parameter.

12. The method according claim 3, which further comprises forming the gradual transition according to a ramp function or a tangens hyperbolicus function.

13. A hearing apparatus, comprising: at least one microphone; and a facility for reducing inherent microphone noise generated independently of ambient noise and receiving signals from said at least one microphone, said facility for reducing said microphone noise having a Wiener filter and an estimation facility coupled to said Wiener filter for determining an estimated value for a noise power, wherein an input signal can be subjected to an attenuation by means of said Wiener filter for generating a processed input signal and a value of the attenuation can be determined on a basis of the estimated value for the noise power, said facility for reducing said microphone noise is set up to monitor the estimated value for the noise power and to deactivate said Wiener filter, if the estimated value is greater than a predetermined limit value.

Description

BRIEF DESCRIPTION OF THE SEVERAL VIEWS OF THE DRAWING

(1) FIG. 1 is a diagrammatic representation of a behind-the-ear hearing device according to the prior art;

(2) FIG. 2 is a block diagram of a facility for reducing microphone noise, which is disposed in a hearing apparatus according to an embodiment of the inventive hearing apparatus;

(3) FIG. 3 is a graph showing a characteristic curve, according to which an attenuation of a Wiener filter of the facility in FIG. 2 is weighted;

(4) FIG. 4 is a block diagram of the hearing apparatus according to a further embodiment of the inventive hearing apparatus;

(5) FIG. 5 is a block diagram showing a signal flow of a beam-former, as can be integrated in the hearing apparatus in FIG. 4;

(6) FIG. 6 is a block diagram of the facility for reducing the microphone noise, as can be provided in the hearing apparatus in FIG. 4;

(7) FIG. 7 is a graph showing a temporal curve of an estimated value for a noise power, as may result in a noise power estimation facility of the hearing apparatus in FIG. 4;

(8) FIG. 8 is a graph showing a setting of maximum attenuation values, as can be provided in the hearing apparatus in FIG. 2 and FIG. 4; and

(9) FIG. 9 is a graph showing a further setting of maximum attenuation values, as can be provided in the hearing apparatus according to FIG. 4.

DETAILED DESCRIPTION OF THE INVENTION

(10) The examples represent preferred embodiments of the invention.

(11) FIG. 2 shows a hearing apparatus 10, which is a behind-the-ear hearing device or an in-the-ear hearing device for instance. A microphone 12 acquires an ambient sound and converts the same into an analog electrical signal, which is converted by a preprocessing facility 14 into a digital input signal x by an analog-digital converter. Provision can also be made with the preprocessing facility 14 to divide the signal of the microphone 12 into a plurality of frequency channels by a filter bank. The input signal x then includes a corresponding number of narrow band partial signals. An output signal is generated from the input signal x by a signal processing facility 16, the output signal being converted by a receiver 18 into a sound signal and being emitted to an ear of a user of the hearing apparatus 10.

(12) The microphone 12 may be a condenser microphone for instance. Aside from the wanted signal (a wanted signal and an ambient noise) generated from the ambient sound, the analog input signal always also contains a microphone noise, which is generated by the microphone 12 itself. In an environment in which it is quiet such that in the input signal x, or at least in one of its frequency channels, the microphone noise has a significantly greater signal power than the signal part generated by the ambient sound, it may nevertheless not result in the user of the hearing apparatus 10 perceiving the microphone noise over the receiver 18. The microphone noise is suppressed by an attenuation W, which in the example shown in FIG. 2, acts as a multiplicative, if necessary frequency-dependent attenuation factor W via a multiplier M on the input signal x or its individual frequency channels. The attenuation factor W is set by a facility 22 for suppressing the microphone noise. If the ambient sound generates a part in the input signal x which is sufficiently large to mask the microphone noise, the attenuation factor W for this time segment and if necessary for the corresponding frequency channel is set by the facility 22 to a value of one or almost one. If the ambient sound is conversely quiet such that the microphone noise may be partially audible by way of the receiver 18, the attenuation factor W is set to a value between zero and one for this period of time and if necessary for the corresponding frequency channel of the attenuation factor W, so that a noticeable attenuation results. The microphone noise is then hereby accordingly reduced in the input signal x.

(13) In order to set the attenuation factor W, the facility 22 contains a facility 24 for calculating a power spectral density (PSD) of the input signal x and a Wiener filter 26 for calculating a gain W. The gain W is calculated by the Wiener filter 26 from the power spectral density PSD of the input signal x and an estimated value for the noise power spectral density (NPSD) according to a function f. The facility 24 may include for instance a simple squaring device for determining an amplitude square of the input signal x or a squaring device and a subordinate smoothing facility for calculating a temporal average value. Every other facility for calculating a power spectral density can also be used here. The function f for calculating the gain W can likewise be a calculation rule which is likewise known per se from the prior art for an attenuation of a noise power contained in a signal. The function f produces a gain W with a value between zero and one, wherein the value aims all the more for one, the greater the ratio shown in FIG. 2 (PSD/NPSD). The function f may also include an estimation of a signal-to-noise ratio (SNR).

(14) The noise power spectral density NPSD is determined by an estimation facility 28 for a noise power contained in the input signal x and from a characteristic curve 30, which describes the typical noise power spectral density of the microphone noise of the microphone 12. The characteristic curve 30 may have been created for instance during the manufacture of the apparatus 10 by measurements. The estimation facility 28 may be a facility which is known per se from the prior art for determining a noise power in a signal.

(15) With the facility 22, a limiter 32, a switch 34 and a masking facility 36 cause the attenuation factor W only to act on the signal parts of the input signal x in which a level is so low that the signal parts are with high probability exclusively or almost exclusively microphone noise from the microphone 12.

(16) As a function of a switch position of the switch 34, either a fixed (frequency-dependent) estimation of the noise power, which was determined on the basis of the characteristic curve 30, or an actual estimation of the noise power from the estimation facility 28, is fed to the Wiener filter 26 and the filtering facility 36. In the event that the estimation facility 28 is used, the estimation of the noise power spectral density NPSD is limited by the limiter 32 to a predetermined highest value. It is assumed for the following explanations that the highest value amounts to 40 dB. With the specification of decibels used here and below, these are decibels for the sound pressure level (SPL). The highest value for the estimation of the noise power can be derived from the characteristic curve 30, wherein an offset of 25 dB for instance can be added to the characteristic curve value.

(17) The combination of the estimation facility 28 and the limiter 32 forms an estimation of the noise power overall, which operates exclusively within the level region of the microphone noise. This causes a value for the gain W to be calculated by the Wiener filter 26 for the function f, the latter automatically striving for one, if the input signal x has a power spectral density which is significantly greater than the highest value of the limiter 32, in other words in this example is greater than 40 dB. With the direct use of a characteristic curve 30 as an estimation for the noise power, as can be achieved by correspondingly switching the switch 24, this produces an automatic deactivation of the Wiener filter 26.

(18) In order additionally to obtain the audio quality of the sound signal of the receiver 18 in the region of levels of the input signal x close to the limitation effected by the limiter 32, the masking facility 36 also produces a gradual transition. The functionality of the masking facility 36 is explained in more detail below with the aid FIG. 3. A diagram is shown for this purpose in FIG. 3, which shows the dependency of a gain factor GF on a current value for the noise power spectral density NPSD. The extent to which the attenuation factor W is calculated from the gain W is determined by the masking facility 36 on the basis of the gain factor FG. A gain factor with the value one signifies W=W. A gain factor with the value zero signifies W=1, i.e. the Wiener filter 26 is deactivated in respect of its influence on the input signal x.

(19) The attenuation factor W is a function of the noise power spectral density NPSD. For a value of the noise power spectral density NPSD<20 dB: W=W applies. For a noise power spectral density NPSDG=30 dB, W=1 applies. A transition 38 is formed there-between by the masking facility 36, which can proceed for instance according to a ramp function 40 or a tangens hyperbolicus function 40. The value G represents a limit value for the activation or deactivation of the Wiener filter.

(20) In order to illustrate the functionality of the facility 22, the expected noise power 42 determined in the diagram by the characteristic curve 30 and the highest value 44 defined by the limiter 32 are shown. The highest value 44 is expediently set equal to the value G, as shown otherwise here.

(21) On the basis of the measurement of the microphone noise, an overall noise level-dependent limitation of the gain W is implemented by the masking facility 36. The closer the estimation of the noise power spectral density NPSD to the limit value G, the more the attenuation is reduced. This ensures that signal parts which are not dominated by microphone noise remain unattenuated. This prevents the facility 22 from interacting with further signal-processing algorithms in the signal processing facility 16. At the same time, the possibility exists of parameterizing the facility 22 in order to suppress the microphone noise independently of the further algorithms, e.g. by a stronger maximum attenuation effected by the gain W being exerted on the microphone noise than on a ambient noise by the signal processing facility 16.

(22) FIG. 4 shows a hearing apparatus 46 with microphones 48, 50, filter banks 52, 54, a beam-former 56, a facility 58 for reducing a microphone noise and a multiplier 60. The digitized microphone signals of the microphone 48 and 50 are combined separately in each instance by the beam-former 56 in individual frequency channels of the filter banks 52, 54 in order to achieve a directional effect. The beam-former 56 may be a beam-former which is known per se from the prior art. By way of example, one possible structure of the beam-former 56 is shown for this purpose in FIG. 5, such as can be provided for processing an individual channel of the filter banks 52, 54. Delay elements with a delay time constant T0 delay the microphone signals of the microphones 48, 50 and are then added to directed signals via an adding device, so that a cardioid signal and an anti-cardioid signal results.

(23) One of the signals is weighted with the value of a directional parameter a by a multiplier, before the two signals are combined to form a directed beam-former signal x by a further adding device. The described arrangement contains a clearly perceivable high pass characteristic. For this reason, low frequencies are amplified by an amplifier 62, in order to render audible for the user the audio information container therein. This amplification also acts on a microphone noise contained in the directed signal x, which is produced by the two microphones 48, 50. On account of the amplification, the microphone noise also contains a different power density spectral distribution in the input signal x for the hearing apparatus 46, which the amplifier 62 generates, from the original microphone noise of the microphones 48, 50 themselves. In addition, the power spectral density of the microphone noise in the input signal x is changed over time by changing the value of the directional parameter a. With the hearing apparatus 46, these properties of the microphone noise are taken into account in the input signal x when calculating an attenuation W, so that a user of the hearing apparatus 46 does not perceive any interfering microphone noise even with a value of the directional parameter a which changes over time.

(24) The input signal x and the directional parameter a form input values for the facility 58. The facility 58, comparable to the facility 22, calculates an attenuation factor W, which acts on the input signal x of the hearing apparatus 46 by way of the multiplier 60. Similarly to the hearing apparatus 10, the attenuation factor W reduces the microphone noise for the input signal x, without in the process a dominating part produced by an ambient sound similarly being influenced in the input signal x by the attenuation factor W.

(25) To explain the mode of operation of the facility 58, this is shown again more precisely in FIG. 6. In FIG. 6 components, which correspond to components in terms of their mode of operation, which are shown in FIG. 2, are provided with the same reference characters as in FIG. 2. They are not shown again in conjunction with FIG. 6.

(26) The change in the power spectral density of the microphone noise effected by the value of the directional parameter a in the input signal x is compensated by the change in the power spectral density being calculated by the calculation facility 64 in the form of a White Noise Gain (WNG) and being taken into account by a divider 66 in the form of a scaling of the input signal x. The noise power spectral density NPSD calculated from the scaled input signal x/WNG by the estimation facility 28 is back-scaled by a multiplier 68 and the value for the White Noise Gain WNG to a back-scaled noise power (NPSD). In conjunction with the beam-former 56 shown in FIG. 5, the following calculation rule is produced as a scaling value WNG for a frequency with a standardized average frequency in the filter banks 52, 54, with which =2**f*Ts and Ts is the scanning time of the analog-to-digital converter of the hearing apparatus 46:
WNG()=[a.sup.2+1+2a*cos(*T0/Ts)]/[1cos(2**T0/Ts)].

(27) With the aid of FIG. 7, the following shows, for an individual frequency component of the input signal x, which estimated values may result for the noise power in the facility 58.

(28) To this end, the lowest diagram in FIG. 7 shows how the value for the directional parameter a is gradually changed with time t during a period of time of 14 seconds, so that an omni-directional directional characteristic of the beam-former 56 results at the start and a notch in the directional characteristic is gradually aligned by a zero steering in different angular directions specified in FIG. 7, in order to then be switched after 12 seconds to an omni-directional directional characteristic. The corresponding value of the White Noise Gain WNG is shown in decibels relating to the corresponding values of the directional parameter a in the graphs above the diagram. For the underlying example in FIG. 7, it should be assumed that the microphone is being operated in a quiet environment, so that the input signal x in the frequency component shown in FIG. 7 exclusively contains the stationary microphone noise. The gradual change in the White Noise Gain (WNG) nevertheless produces a curve of the sum square |x|.sup.2 of the input signal x, as is shown in the topmost diagram in FIG. 7. On the basis of this curve, the estimation facility 28 would, on account of its inertia, not be in a position to correctly reproduce the noise power at the transition points (e.g. at second 2). The scaling by the divider 66 produces, as an input signal for the estimation facility 28, a stationary input x/WNG during the curve. The estimation facility 28 calculates a correspondingly correct noise power spectral density NPSD relating to the scaled input signal x/WNG. The back-scaling by the multiplier 68 then produces a correspondingly correct estimation NPSD for the actual noise power contained in the input signal x. This is used to calculate a gain W suited to effectively attenuating the microphone noise by the Wiener filter 26.

(29) With the facility 58, the further components explained in conjunction with FIG. 2 can be provided to deactivate the Wiener filter 26 by limiter, a switch and a masking facility. These components are not shown again for the sake of clarity in FIG. 6.

(30) FIGS. 8 and 9 show two alternative options, which further improve the audio quality of the input signal x processed by the multiplier 20 or 60. The diagrams are shown here in the instance that a beam-former, like the beam-former 56, is used. The noise power of the microphone noise, in particular for low frequencies, can be significantly amplified by the beam-former 56 as a function of the value for the directional parameter a in the ratio of the original microphone noise of the microphone 48, 50. FIGS. 8 and 9 to this end show the level of the microphone noise for a specific time instant for several channels C of the filter banks 52, 54 and a specific setting of the parameter a prior to attenuation by the multiplier 60 (as a bar chart |x|.sup.2) and after attenuation (as bar chart |x|.sup.2*W.sup.2). In order not to amplify a fluctuation in the level of the microphone noise on account of the time-variant attenuation M, the attenuation W is restricted to a maximum attenuation value NF, which amounts here logarithmically in the example shown in FIG. 8 NF=10 dB. Accordingly, signal parts of the microphone noise in the input signal remain perceivable to a user in the low-frequency range (here in particular the channels 0 to 7) even after attenuation. These nevertheless contain less interfering modulation on account of limiting the attenuation to the maximum attenuation value NF.

(31) A strong attenuation of the microphone noise of this type, which is no longer perceivable to the user him/herself, is produced for the remaining channels (channels C=6-47). The microphone noise also has stationary behavior after the processing, which it has also featured prior to the processing by the beam-former.

(32) In order also to reach the maximum attenuation NF when determining the value, such that the microphone noise is reduced to a comfortable level, the beam-former characteristic, e.g. in the form of the value of the directional parameter a, can also be taken into account. A frequency-dependent maximum attenuation NF(C, a) can be determined by the White Noise Gain WNG. The aim here is to achieve an attenuated microphone noise, in which the channels C have an almost identical level of the microphone noise and this level is independent of a momentary setting of the beam-former, i.e. the value for the directional parameter a.

(33) Such a frequency-dependent setting of the maximum attenuation NF(C,a) is shown in FIG. 9 in the right-hand diagram. The values for the maximum attenuation NF(C,a) are frequency and also time-dependent and represent a function of the value of the directional parameter a. The left diagram shows how a spectrally almost flat course of the microphone noise is achieved by a maximal limitation NF(C, a) of this type. Limiting the attenuation to a maximum attenuation can be implemented within the Wiener filter 26 for instance. It should be noted here that limiting the attenuation means that the Wiener gain W does not become smaller than a value corresponding to the value NF or NF(C, a). Limiting the attenuation factor W to small values produces a so-called noise floor in the processed input signal.

(34) On the basis of the value for the directional parameter a, the limit value G can also be set for the masking facility 36, if this is provided in the facility 58. This herewith then prevents the Wiener filter from deactivating because a level of the microphone noise results on account of the beam-former 56, which is greater than the level of the microphone noise to be expected on account of the characteristic curve 30.

(35) In summary, it should be noted that with a beam-former with an adjustable directional characteristic, an efficient reduction in the microphone noise is possible to a comfortable level. In addition, the approach is advantageous in that so-called noise flags are prevented, which are otherwise typically caused in a signal of a beam-former. Such noise flags may follow a signal of an external sound source, such as for instance a speaker, if this sound source falls silent and the microphone noise is then audible for the user of the hearing apparatus, because it is not attenuated sufficiently quickly. The rapid adjustment is enabled with the approaches inter alia by the limiter 32, which keeps the estimation of the noise power of the microphone noise NPSD to a level which already lies very close to the actual microphone noise.