EP1451813B1 - Method for suppressing surrounding noise in a hands-free device, and hands-free device - Google Patents

Abstract

In order to suppress as much surrounding noise as possible in a hands-free device in a motor vehicle for example, two microphones (M1, M2) are arranged at a distance from each other. The output signals (MS1, MS2) of said microphones are added in an adder (AD) and subtracted in a subtracter (SU). The adder's (AD) cumulative signal (S) is Fourier transformed in a first Fourier transformer (F1), and the subtracter's (SU) differential signal (D) is Fourier transformed in a second Fourier transformer (F2). A speech pause detector (P) detects speech pauses from the two Fourier transformed R(f) and D(f) during which a third arithmetic unit (R) calculates the transfer function HT of an adaptable transformation filter (TF) from the spectral power density Srr of the cumulative signal (S) and the spectral power density SDD of the differential signal (D). The transfer function of a spectral subtraction filter (SF), at the inlet of which the cumulative signal's (S) Fourier transformed R(f) is located, is created from the spectral power density Srr of the cumulative signal (S) and the interference power density Snn generated by the adaptable transformation filter (TF). The outlet of the spectral subtraction filter (SF) is connected to the inlet of an inverse Fourier transformer (IF), at the outlet of which an audio signal (A) can be picked up in the time domain that is virtually free of surrounding noise.

Description

The invention relates to a method for suppressing ambient noise in a speakerphone with two microphones arranged at a predeterminable distance.

The invention further relates to a hands-free device having two microphones arranged at a predeterminable distance from one another.

When using hands-free equipment, ambient noise is a major disruptive factor that can significantly affect speech intelligibility. Car telephones are equipped with hands-free equipment so that the driver can fully concentrate on driving the vehicle and on the road. In a vehicle but especially loud and disturbing ambient noise occur.

The dissertation Stefan Gierl: "Noise reduction in speech transmission with the help of microphone array systems", University of Karlsruhe, 1990 describes a microphone array system for noise reduction in speech transmission, which is capable of suppressing both coherent and incoherent noise. By applying a spectral subtraction method, speech-canceled interference references are calculated from balanced microphone signals, from which in turn estimated values for the interference power density spectra are determined and permanently updated. On speech pause detectors can be dispensed with.

It is an object of the invention to design a method for suppressing ambient noise for a hands-free device and a speakerphone so that ambient noise can be suppressed as completely as possible, without the need mehrkanälige microphone arrays are necessary.

The method, this object is achieved with the features specified in claim 1.

In terms of apparatus, this object is achieved with the features specified in claim 10.

The hands-free device according to the invention is equipped with two microphones, which are arranged at a predeterminable distance from one another. The distance of the speaker to the microphones is chosen smaller than the so-called. Hall radius, so that the direct sound components of the speaker at the location of the microphones dominate over the reflection components occurring in space. From the microphone signals supplied by the microphones, the sum and the difference signal are formed, from which the Fourier transform of the sum and the Fourier transform of the difference signal are formed by means of a respective Fourier transformer.

From these Fourier transform the speech pauses z. B. detected by the fact that their average short-term performance are determined. During speech pauses, the short-term powers of the sum and difference signals are approximately equal, because with uncorrelated signal components, it does not matter if they are added or subtracted prior to the power calculation, while at the beginning of speech, due to the highly correlated speech component, the short-term power in the sum signal is greater than the short-term power in the difference signal increases significantly. This increase can be easily detected and used for the reliable detection of a speech break. A speech break can therefore be detected with great certainty even in the case of loud ambient noise.

The method according to the invention further provides for determining the spectral power density from the Fourier transform of the sum signal and from the Fourier transform of the difference signal, from which the transfer function for an adaptive transformation filter is calculated. This adaptive transformation filter generates the interference power density by multiplying the power density of the Fourier transform of the difference signal by its transfer function. From the spectral power density of the Fourier transform of the sum signal and from the interference power density generated by the adaptive transform filter, the transfer function of a likewise adaptive Spektralsubttraktionsfilters is calculated, which filters the Fourier transform of the sum signal and provides at its output a largely free from ambient noise audio signal in the frequency domain, by means of an inverse Fourier transformer is transformed back into the time domain. At the output of this inverse Fourier transformer can therefore a largely of ambient noise free audio or voice signal in the time domain removed and further processed.

The inventive method and the hands-free device according to the invention are described and explained in more detail with reference to the embodiment shown in the figure.

The output of a first microphone M1 is connected to the first input of an adder AD and the first input of a subtractor SU, while the output of a second microphone M2 is connected to the second input of the adder AD and the second input of the subtractor SU. The output of the adder AD is connected to the input of a first Fourier transformer F1, whose output to the first input of a speech pause detector P, the input of a first arithmetic unit LS for calculating the spectral power density S _{rr of} the Fourier transform R (f) of the sum signal S and Input of an adaptive spectral subtraction filter SF is connected.

The output of the subtractor SU is connected to the input of a second Fourier transformer F2 whose output is connected to the second input of the speech pause detector P and to the input of a second arithmetic unit LD for calculating the spectral power density S _{DD of} the Fourier transform D (f) of the difference signal D. , The output of the first arithmetic unit LS is connected to a third arithmetic unit for calculating the transfer function of an adaptive transformation filter TF and to the first control input of the adaptive spectral subtraction filter SF whose output is connected to the input of an inverse Fourier transformer IF. The output of the second arithmetic unit LD is connected to the third arithmetic unit R and to the input of the adaptive transformation filter TF whose output is connected to the second control input of the adaptive spectral subtraction filter SF. The output of the speech pause detector P is also connected to the third arithmetic unit R whose output is connected to the control input of the adaptive transformation filter TF.

The two microphones M1 and M2 are, as already mentioned, arranged at a distance from the speaker, which is smaller than the so-called Hall radius. For this reason, the direct sound components of the speaker at the location of the microphones dominate over those in a closed space, eg. B. the interior of a vehicle, occurring reflection components.

The sum signal S of the microphone signals MS1 and MS2 of the two microphones M1 and M2 is formed in the adder AD, while the difference signal D of the microphone signals MS1 and MS2 is formed in the subtracter SU.

The first Fourier transformer F1 forms the Fourier transform R (f) of the sum signal S. Likewise, the second Fourier transformer F2 forms the Fourier transform D (f) of the difference signal D.

In the speech pause detector P, the short-time power of the Fourier transform R (f) of the sum signal S and the Fourier transform D (f) of the difference signal D is determined. During speech pauses, the two short-term performances hardly differ from each other, because it does not matter for uncorrelated signal components whether they are added or subtracted before the power calculation. At the beginning of speech, on the other hand, due to the strongly correlated speech component, the short-time power in the sum signal increases significantly compared with the short-time power in the difference signal. This increase therefore indicates the end of a speech break and the beginning of speech.

The first arithmetic unit LS calculates by time averaging the spectral power density S _{rr of} the Fourier transform R (f) of the sum signal S. Similarly, the second arithmetic unit LD calculates the spectral power density S _{DD of} the Fourier transform D (f) of the difference signal D. The third arithmetic unit R calculates now from the power density S _rrp (f) and from the spectral power density S _DDp (f) during the speech pauses the transfer function H _T (f) of the adaptive transformation filter TF according to the following formula (1) $${\mathrm{H}}_{\mathrm{T}}\left(\mathrm{f}\right)\mathrm{=}{\mathrm{S}}_{\mathrm{rrp}}\left(\mathrm{f}\right)\mathrm{/}{\mathrm{S}}_{\mathrm{DDp}}\left(\mathrm{f}\right)$$

Preferably, by means of an additional temporal averaging-that is to say a smoothing-of the coefficients of the transfer function obtained in this way, the suppression of environmental noises is considerably improved because the occurrence of so-called artifacts, which are frequently also referred to as "musical tones", is prevented.

The spectral power density S _rr (f) is obtained by temporal averaging of the Fourier transform R (f) of the sum signal S, while in the same way the spectral power density S _DD (f) calculated by time averaging from the Fourier transform D (f) of the difference signal D. becomes.

For example, the spectral power density S _{rr is} calculated according to the following formula (2): $${\mathrm{S}}_{\mathrm{rr}}\left(\mathrm{f},\mathrm{k}\right)\mathrm{=}\mathrm{c}\mathrm{*}{\left|\mathrm{R}\left(\mathrm{f}\right)\right|}^{\mathrm{2}}\mathrm{+}\left(\mathrm{1}\mathrm{-}\mathrm{c}\right)\mathrm{*}{\mathrm{S}}_{\mathrm{rr}}\left(\mathrm{f}\mathrm{.}\mathrm{k}\mathrm{-}\mathrm{1}\right)$$

Analogously, z. B. the spectral power density S _DD (f) calculated according to the following formula (3): $${\mathrm{S}}_{\mathrm{DD}}\left(\mathrm{f},\mathrm{k}\right)\mathrm{=}\mathrm{c}\mathrm{*}{\left|\mathrm{D}\left(\mathrm{f}\right)\right|}^{\mathrm{2}}\mathrm{+}\left(\mathrm{1}\mathrm{-}\mathrm{c}\right)\mathrm{*}{\mathrm{S}}_{\mathrm{DD}}\left(\mathrm{f}\mathrm{.}\mathrm{k}\mathrm{-}\mathrm{1}\right)$$

c is a constant lying between 0 and 1, which determines the averaging period. For c = 1 no time averaging takes place, but instead the squares of the Fourier transforms R (f) and D (f) are taken directly as estimates for the spectral power densities. The calculation of the remaining spectral power densities, which are required for carrying out the method according to the invention, preferably takes place in the same way.

The adaptive transformation filter TF generates H _T (f) from the spectral power density by means of its transfer function S _DD (f) of the Fourier transform D (f) the interference power density S _nn according to the following formula (4): $${\mathrm{S}}_{\mathrm{nn}}\left(\mathrm{f}\right)\mathrm{=}{\mathrm{H}}_{\mathrm{T}}\mathrm{*}{\mathrm{S}}_{\mathrm{DD}}\left(\mathrm{f}\right)$$

With the aid of the interference power density S _nn calculated from the Fourier transform D (f) of the difference signal D and the spectral power density S _{rr of} the sum signal, ie of the disturbed signal, calculated by the first arithmetic unit LS, the transfer function H _{sub of} the spectral subtraction filter SF is determined according to the following rule (5). calculated: $$\begin{array}{lll}{\mathrm{H}}_{\mathrm{sub}}\left(\mathrm{f}\right)& \mathrm{=}\mathrm{1}\mathrm{-}\mathrm{a}\mathrm{*}{\mathrm{S}}_{\mathrm{nn}}\left(\mathrm{f}\right)\mathrm{/}{\mathrm{S}}_{\mathrm{rr}}\left(\mathrm{f}\right)& \mathrm{f}\ddot{\mathrm{u}}\mathrm{r}\mathrm{1}\mathrm{-}\mathrm{a}\mathrm{*}{\mathrm{S}}_{\mathrm{nn}}\left(\mathrm{f}\right)\mathrm{/}{\mathrm{S}}_{\mathrm{rr}}\left(\mathrm{f}\right)\mathrm{>}\mathrm{b}\\ {\mathrm{H}}_{\mathrm{sub}}\left(\mathrm{f}\right)& =\mathrm{b}& \mathrm{f}\ddot{\mathrm{u}}\mathrm{r}\mathrm{1}\mathrm{-}\mathrm{a}\mathrm{*}{\mathrm{S}}_{\mathrm{nn}}\left(\mathrm{f}\right)\mathrm{/}{\mathrm{S}}_{\mathrm{rr}}\left(\mathrm{f}\right)\le \mathrm{b}\end{array}$$

The parameter a represents the so-called overestimation factor, while b represents the so-called "spectral floor".

The noise components picked up by the microphones M1 and M2, which strike the microphones M1 and M2 as diffuse sound waves, can be considered to be almost uncorrelated for almost the entire frequency band of interest. However, depending on the distance between the two microphones M1 and M2 to each other at low frequencies, there is still a certain correlation, which leads to the fact that the interference components contained in the reference signal appear to be high-pass filtered. In order to avoid a misjudgment of the low-frequency interference components during spectral subtraction, a spectral increase of the low-frequency components of the reference signal takes place with the aid of the adaptive transformation filter TF shown in the FIGURE.

The inventive method and the hands-free circuit according to the invention, which are particularly suitable for a car phone, are characterized by an excellent voice quality and speech intelligibility, because the estimated value for the noise power density Snn irrespective of the voice activity is permanently updated. Thus, also the transfer function of the spectral subtraction filter SF is constantly updated, both during voice activity and during the speech pauses. As already mentioned, speech pauses are reliably and accurately detected, which is necessary for updating the transformation filter TF.

The audio signal at the output of the spectral subtraction filter SF, which is largely free from ambient noise, is fed to an inverse Fourier transformer IF, which transforms the audio signal back into the time domain.

LIST OF REFERENCE NUMBERS

A

in the time domain transformed back audio signal

AD

adder

D

difference signal

D (f)

Fourier transform of the difference signal

F1

first Fourier transformer

F2

second Fourier transformer

H _sub

Transfer function of the spectral subtraction filter

H _T

Transfer function of the transformation filter

IF

inverse Fourier transformer

LD

second arithmetic unit for calculating the spectral power density

LS

first arithmetic unit for calculating the spectral power density

MS1

microphone signal

MS2

microphone signal

M1

microphone

M2

microphone

P

Speech pause detector

R

third arithmetic unit for calculating the transfer function of the transformation filter

R (f)

Fourier transform of the sum signal

S

sum signal

SF

Spektralsubtraktionsfilter

SU

subtractor

S _DD

spectral power density of the difference signal

S _nn

interference power density

S _rr

spectral power density of the sum signal

TF

transform filter

Claims (18)

1. A method of suppressing ambient noise in a hands-free device with two microphones (M1, M2) at a predeterminable distance from one another, which supply a respective microphone signal (MS1, MS2) with the following method steps:

   the sum signal (S) and the difference signal (D) of the two microphone signals (MS1, MS2) are formed,

   the Fourier transform R(f) of the sum signal (S) and the Fourier transform D(f) of the difference signal (D) are formed,

   gaps in conversation are detected from the Fourier transforms R(f) and D(f),

   the spectral power density S_n is determined from the Fourier transform R(f) of the sum signal (S),

   the spectral power density S_DD is determined from the Fourier transform D(f) of the difference signal (D),

   during the gaps in conversation the transmission function H_T(f) for an adaptive transformation filter (TF) is calculated from the spectral power density S_n of the Fourier transform R(f) of the sum signal (S) and from the spectral power density S_DD of the Fourier transform D(f) of the difference signal (D),

   the adaptive transformation filter (TF) produces the interference power density S_nn(f) by multiplication of the power density S_DD of the Fourier transform D(f) of the difference signal (B) by its transmission function H_T(f),

   the transmission function H_sub(f) of a spectral subtraction filter (SF) is calculated from the interference power density S_nn(f) and from the spectral power density S_n of the Fourier transform R(f) of the sum signal (S),

   the spectral subtraction filter (SF) filters the Fourier transform R(f) of the sum signal (S) and

   the output signal of the spectral subtraction filter (SF) is transformed back into the time domain.
2. A method as claimed in claim 1, characterised in that the transmission function H_T(f) of the transformation filter (TF) during gaps in conversation is given by the following formula: $${\mathrm{H}}_{\mathrm{T}}\left(\mathrm{f}\right)\mathrm{=}{\mathrm{S}}_{\mathrm{rrP}}\left(\mathrm{f}\right)\mathrm{/}{\mathrm{S}}_{\mathrm{DDP}}\left(\mathrm{f}\right)$$

   
3. A method as claimed in claim 2, characterised in that the coefficient of the transmission function H_T(f) of the transformation filter (TF) is averaged over time.
4. A method as claimed in claim 1, 2 or 3, characterised in that the calculation of the spectral power density S_rr from the Fourier transform R(f) of the sum signal (S) and the spectral power density S_DD from the Fourier transform D(f) of the difference signal (D) is effected by averaging over time.
5. A method as claimed in claim 4, characterised in that the spectral power density S_rr is calculated in accordance with the following formula: $${\mathrm{S}}_{\mathrm{rr}}\left(\mathrm{f},\mathrm{k}\right)\mathrm{=}\mathrm{c}\mathrm{*}{\left|\mathrm{R}\left(\mathrm{f}\right)\right|}^{\mathrm{2}}\mathrm{+}\left(\mathrm{1}\mathrm{-}\mathrm{c}\right)\mathrm{*}{\mathrm{S}}_{\mathrm{rr}}\left(\mathrm{f}\mathrm{,}\mathrm{k}\mathrm{-}\mathrm{1}\right)$$

   

   
   wherein k represents a time index and c is a constant for determining the averaging duration.
6. A method as claimed in claim 4 or 5, characterised in that the spectral power density S_DD is calculated in accordance with the following formula: $${\mathrm{S}}_{\mathrm{DD}}\left(\mathrm{f},\mathrm{k}\right)\mathrm{=}\mathrm{c}\mathrm{*}{\left|\mathrm{D}\left(\mathrm{f}\right)\right|}^{\mathrm{2}}\mathrm{+}\left(\mathrm{1}\mathrm{-}\mathrm{c}\right)\mathrm{*}{\mathrm{S}}_{\mathrm{DD}}\left(\mathrm{f}\mathrm{,}\mathrm{k}\mathrm{-}\mathrm{1}\right)$$

   

   
   wherein k represents a time index and c is a constant for determining the averaging duration.
7. A method as claimed in one of claims 1 to 6, characterised in that the short term power of the Fourier transform R(f) of the sum signal (S) and of the Fourier transform D(f) of the difference of signal (D) is determined for the detection of the gaps in conversation and that a gap in conversation is detected when the two determined short term powers lie within a predeterminable, common tolerance range.
8. A method as claimed in one of claims 1 to 7, characterised in that the transmission function H_sub(f) of the spectral subtraction filter (SF) is calculated in accordance with the following formula: $$\begin{array}{l}{\mathrm{H}}_{\mathrm{sub}}\left(\mathrm{f}\right)=\mathrm{1}\mathrm{-}\mathrm{a}\mathrm{*}{\mathrm{S}}_{\mathrm{nn}}\left(\mathrm{f}\right)\mathrm{/}{\mathrm{S}}_{\mathrm{rr}}\left(\mathrm{f}\right)\mathrm{for}\mathrm{1}\mathrm{-}\mathrm{a}\mathrm{*}{\mathrm{S}}_{\mathrm{nn}}\left(\mathrm{f}\right)\mathrm{/}{\mathrm{S}}_{\mathrm{rr}}\left(\mathrm{f}\right)\mathrm{>}\mathrm{b}\\ {\mathrm{H}}_{\mathrm{sub}}\left(\mathrm{f}\right)=\mathrm{b\; for}\mathrm{1}\mathrm{-}\mathrm{a}\mathrm{*}{\mathrm{S}}_{\mathrm{nn}}\left(\mathrm{f}\right)\mathrm{/}{\mathrm{S}}_{\mathrm{rr}}\left(\mathrm{f}\right)\le \mathrm{b}\\ \end{array}$$

   

   
   wherein a represents the so-called `overestimation factor' and b represents the so-called `spectral floor'.
9. A method as claimed in one of claims 1 to 8, characterised in that the delay time differences between the two microphone signals (MS1, MS2) are balanced out.
10. A hands-free device with two microphones (M1, M2) arranged at a predeterminable spacing from one another, characterised in that the output of the first microphone (M1) is connected to the first input of an adder (AD) and the first input of a subtractor (SU),
    that the output of the second microphone (M2) id connected to the second input of the adder (AD) and the second input of the subtractor (SU),
    that the output of the adder (AD) is connected to the input of a first Fourier transformer (F1), the output of which is connected to the first input of a conversational gap detector (P), to the input of a first computing unit (LS) for calculating the spectral power density S_rr and to the input of an adaptive spectral subtraction filter (SF),
    that the output of the subtractor (SU) is connected to the input of a second Fourier transformer (F2), whose output is connected to the second input of the conversational gap detector (P) and to the input of a second computing unit (LD) for calculating the spectral power density S_DD,
    that the outputs of the conversational gap detector (P), of the first computing unit (LS) and of the second computing unit (LD) are connected to a third computing unit (R) for calculating the transmission function H_T(F) of an adaptive transformation filter (TF) during detected gaps in conversation,
    that the output of the first computing unit (LS) is connected to the first control input of the adaptive spectral subtraction filter (SF),
    that the output of the third computing unit (R) is connected to the control input of the adaptive transformation filter (TF), the input of which is connected to the output of the second computing unit (LD) and the output of which is connected to the second control input of the adaptive spectral subtraction filter (SF), and
    that the output of the adaptive spectral subtraction filter (SF) is connected to the input of an inverse Fourier transformer (IF), at whose output an audio signal (A) transformed back into the time domain may be taken off.
11. A hands-free device as claimed in claim 10, characterised in that the transmission function H_T(f) of the transformation filter (TF) during gaps in conversation is formed in accordance with the following formula: $${\mathrm{H}}_{\mathrm{T}}\left(\mathrm{f}\right)\mathrm{=}{\mathrm{S}}_{\mathrm{rrp}}\left(\mathrm{f}\right)\mathrm{/}{\mathrm{S}}_{\mathrm{DDp}}\left(\mathrm{f}\right)$$

    
12. A hands-free device as claimed in claim 11, characterised in that the coefficients of the transmission function H_T(f) of the transformation filter (TF) are averaged over time.
13. A hands-free device as claimed in claim 10, 11 or 12, characterised in that the spectral power density S_rr is formed from the Fourier transform R(f) of the sum signal (S) and that the spectral power density S_DD is formed from the Fourier transform D(f) of the difference signal (D) by averaging over time.
14. A hands-free device as claimed in claim 13, characterised in that the spectral power density S_rr is calculated in accordance with the following formula: $${\mathrm{S}}_{\mathrm{rr}}\left(\mathrm{f},\mathrm{k}\right)\mathrm{=}\mathrm{c}\mathrm{*}{\left|\mathrm{R}\left(\mathrm{f}\right)\right|}^{\mathrm{2}}\mathrm{+}\left(\mathrm{1}\mathrm{-}\mathrm{c}\right)\mathrm{*}{\mathrm{S}}_{\mathrm{rr}}\left(\mathrm{f}\mathrm{,}\mathrm{k}\mathrm{-}\mathrm{1}\right)$$

    

    
    wherein k represents a time index and c is a constant for determining the averaging duration.
15. A hands-free device as claimed in claim 13 or 14, characterised in that the spectral power density S_DD is calculated in accordance with the following formula: $${\mathrm{S}}_{\mathrm{DD}}\left(\mathrm{f},\mathrm{k}\right)\mathrm{=}\mathrm{c}\mathrm{*}{\left|\mathrm{D}\left(\mathrm{f}\right)\right|}^{\mathrm{2}}\mathrm{+}\left(\mathrm{1}\mathrm{-}\mathrm{c}\right)\mathrm{*}{\mathrm{S}}_{\mathrm{DD}}\left(\mathrm{f}\mathrm{,}\mathrm{k}\mathrm{-}\mathrm{1}\right)$$

    

    
    wherein k represents a time index and c is a constant for determining the averaging duration.
16. A hands-free device as claimed in one of claims 10 to 15, characterised in that the short term power of the Fourier transform R(f) of the sum signal (S) and of the Fourier transform D(f) of the difference signal (D) is determined in order to detect the gaps in conversation and that a gap in conversation is detected when the two determined short term powers lie within a predeterminable, common tolerance range.
17. A hands-free device as claimed in one of claims 10 to 16, characterised in that the transmission function H_sub(f) of the spectral function filter (SF) is calculated in accordance with the following formula: $$\begin{array}{l}{\mathrm{H}}_{\mathrm{sub}}\left(\mathrm{f}\right)=\mathrm{1}\mathrm{-}\mathrm{a}\mathrm{*}{\mathrm{S}}_{\mathrm{nn}}\left(\mathrm{f}\right)\mathrm{/}{\mathrm{S}}_{\mathrm{rr}}\left(\mathrm{f}\right)\mathrm{for}\mathrm{1}\mathrm{-}\mathrm{a}\mathrm{*}{\mathrm{S}}_{\mathrm{nn}}\left(\mathrm{f}\right)\mathrm{/}{\mathrm{S}}_{\mathrm{rr}}\left(\mathrm{f}\right)\mathrm{>}\mathrm{b}\\ {\mathrm{H}}_{\mathrm{sub}}\left(\mathrm{f}\right)=\mathrm{b\; for}\mathrm{1}\mathrm{-}\mathrm{a}\mathrm{*}{\mathrm{S}}_{\mathrm{nn}}\left(\mathrm{f}\right)\mathrm{/}{\mathrm{S}}_{\mathrm{rr}}\left(\mathrm{f}\right)\le \mathrm{b}\\ \end{array}$$

    

    
    wherein a represents the so-called 'overestimation factor' and b represents the so-called 'spectral floor'.
18. A hands-free device as claimed in one of claims 10 to 17, characterised in that the delay time differences between the two microphone signals (M1, M2) may be balanced out.

Images