Wind detection method

Abstract

Wind detection method The invention relates to a wind detection method implemented in a sound reproduction device and comprising the following steps: - measuring a first external signal sampled from a first external microphone (30), - measuring a second external signal sampled from a second external microphone (35), and - calculating a cross-correlation between the first and second external signals further comprises: - calculating a variation over time of a phrase of the cross-correlation, and - determining the presence of wind as a function at least of the calculated phase variation. Figure for abstract: Figure 1

Description

Title of the invention: Wind detection method

[0001] The present invention relates to a wind detection method, of the type implemented in a sound reproduction device and comprising the following steps:

[0002] - the measurement of a first sampled external signal (X[n]) from a first external microphone;

[0003] - the measurement of a second sampled external signal (K[n]) from a second mi external microphone; and

[0004] - calculating a cross-correlation (XFfrz]) between the first and second signals external FM).

[0005] Earpieces or headphones often incorporate noise-cancelling treatment to attenuate or amplify external noise.

[0006] The noise reduction processing chains use an external microphone to capture ambient sound outside the cavity.

[0007] When a microphone is in an airflow, the pressure variation from the turbulence of that airflow deflects the microphone diaphragm, resulting in wind noise in the microphone output signal.

[0008] The external microphone which is used for active noise cancellation feeds the noise reduction controller with a noisy signal which requires specific processing taking into account the existence or not of wind.

[0009] The implementation of wind detection means is described in document US 2004 / 0161120 AL

[0010] This document relates to a device for detecting the presence of wind comprising two external microphones. Each microphone is connected to a processing device which generates a cross-correlation between a first and a second signal from the microphones. The processing device also generates a signal corresponding to the autocorrelation of the first or second signal.

[0011] The processing device compares the values of the autocorrelation and the cross-correlation. The comparison means are arranged to detect whether the autocorrelation of a signal is greater than the cross-correlation of the two signals. Satisfaction of this condition indicates the presence of wind noise.

[0012] These wind detection means are relatively unreliable for wind detection.

[0013] The aim of the invention is to propose a more reliable wind detection method.

[0014] To this end, the invention relates to a method for detecting wind, of the aforementioned type, characterized in that it further comprises:

[0015] - the calculation of a variation ( A 4nl) over the time of a sentence ( <p[w]) de la cor cross-relation (XX[ n ] ); and

[0016] - determining the presence of wind as a function of at least the variation of calculated phase (A ^z])-

[0017] According to particular embodiments, the method comprises one or more of the following characteristics: - the calculation of an energy (EnFF[n]) of at least one external signal and the determination of the presence of wind as a function of at least the calculated phase variation (A <4Xb and the calculated energy (EnFF[n]); - the calculation of an auto-correlation (XX [ n ], YX[ / î]) of each first and second sampled external signal (X[ft], a calculation of coherence (C between the two calculated autocorrelations and the determination of the presence of wind as a function at least of the calculated phase variation (A 44]) and of the calculated coherence (£ ; - a step of calculating a wind estimator (VE£[ n]) equal to a phase variation estimator dependent on the calculated phase variation (A <44]) - the product of a coherence estimator dependent on the calculated coherence (C [nï], an energy estimator dependent on the calculated energy (EnFF[ri$) and the phase variation estimator, and in which the determination of the presence of wind comprises the comparison of the wind estimator (PXE[n]) with at least one predetermined threshold (Tl); - the wind estimator WE [n] is given by the following formula:

[0018] W£[„] = 1-{1_ [„])"'■ ■ (1- A?[<' OR

[0019] Cwean[n] is the calculated coherence;

[0020] A (p [ n ] is the phase variation;

[0021] EnFF[ n] is the energy of at least one of the first and second signals; and

[0022] a1, a2, and a3 are weights; - calculating the cross-correlation (XX [ n ] ) recursively so that the cross-correlation at the next step is calculated based on the signals sampled at the next step (X[n], X[n]) and the cross-correlation ( XY [ n - 1 ] ) at the previous step; - the calculation of the cross-correlation XY [ n ] is given by the formula:

[0023] XK[n] -a • XX[n-1] + (1-a) • X[n] • conj(X[n])

[0024] Where: a is a parameter between 0 and 1, X[n] and are transforms of Fourier of the signals measured from the first and second microphones (30, 35),

[0025] conj(X[n]) is the conjugate of X[n] and XY[n-1] is the cross-correlation of the two signals at the previous step.

[0026] The invention further relates to a method for processing external noise comprising: - detecting wind by implementing a method such as above, and applying external noise processing dependent on wind detection.

[0027] According to particular embodiments, the method comprises one or more of the following characteristics: - the external noise processing includes a noise reduction mode (ANC) in which, in the presence of wind, the signal from at least one of the external microphones is filtered by a peak filter at a frequency between 1.5 and 2.5 kHz; and - the external noise processing includes a transparent mode in which, in the presence of wind, the signal from at least one of the external microphones is filtered both by a stepped low pass at a frequency between 200 and 300 Hz and a peak filter at a frequency between 800 Hz and 1 kHz.

[0028] The invention also relates to a wind detection device comprising:

[0029] - a first and a second external microphones and their own processing means to the implementation of the method as defined above. The invention also relates to a sound reproduction device comprising; - - it includes a first and a second external microphones and a module of external noise processing suitable for implementing a method such as above.

[0030] The invention will be better understood on reading the description which follows, given solely by way of example and made with reference to the drawings in which: - [Fig.l] [Fig.l] is a schematic view of an earpiece according to the invention; - [Fig.2] [Fig.2] is a block diagram of the earpiece according to the invention; and - [Fig.3] [Fig.3] is a block diagram of the wind detector according to the invention.

[0031] In [Fig.l], the earpiece 1 comprises, as known, a tip 5 suitable for being introduced into the ear canal. This tip delimits a cavity 10 for sound reproduction. A housing 15 extends the cavity 10 outside the ear. This housing 15 receives, as known per se, the electronic components of the earpiece.

[0032] An electroacoustic transducer 20 is arranged in this cavity 10 opposite the auditory canal of the ear. This transducer 20 is capable of emitting either an anti-noise signal or a transparency signal in the cavity 10 and possibly also of ensuring the reproduction of a sound signal such as music or voice.

[0033] The transducer 20 is connected to an external noise processing module 25 for receive an excitation signal.

[0034] The earpiece comprises a first external microphone 30 used only for noise attenuation. The earpiece also comprises a second microphone 35 suitable for picking up the voice of the wearer of the earpiece when he speaks.

[0035] The two microphones 30, 35 are each capable of picking up an ambient sound outside the cavity 10 according to distinct orientations. The two microphones 30, 35 are connected to a wind detector 40.

[0036] In addition, the earpiece comprises an internal microphone 45 disposed in the cavity 10 for capturing internal sound in the cavity.

[0037] The internal microphone 45 is connected to the external noise processing module 25 to provide a feedback loop.

[0038] In [Fig.2], the earpiece elements illustrated in [Fig.l] bear the same reference number. Thus, we find the two external microphones 30 and 35 as well as the internal microphone 45 and the wind detector 40.

[0039] The external noise processing module 25 comprises, at the input, a synthesis unit 50 to which the two external microphones 30 and 35 are connected. This synthesis unit 50 is capable of combining the two signals from the external microphones 30 and 35 into a single signal, taking into account the position of the two microphones on the earpiece and their orientation to reconstitute a single signal representative of the external sound perceived in a single direction. The combination of these two microphones makes it possible, for example, to form a beam (beamforming in English) in the direction of the user's mouth, in the case of a voice recording. The direction of the formed beam is adjustable, via the value of the delay applied to the signals between the two microphones.

[0040] The signal from the synthesis unit 50 is sent to an adaptive wind noise reduction filter 55. This wind noise reduction filter is adaptive depending on whether or not wind has been detected and depending on the type of processing provided by the external noise processing module 25.

[0041] For this purpose, the wind noise detector 40 is connected to the filter 55 to address a signal representative of the presence of wind, a value 0 if there is no wind, and a value 1 if there is wind for example.

[0042] The earpiece comprises a manual selector 56 for selecting the operating mode. The two operating modes are the transparent mode which allows external noise to be heard without being attenuated by the structure of the earpiece and the noise reduction mode which attenuates external noise as much as possible. The switch 56 is connected to the wind reduction filter 55 so that it adapts the filter implemented, in the presence of wind according to the selected operating mode.

[0043] If the noise attenuation mode is chosen, the wind reduction filter 55 puts in implements a filter in which the signal from at least one of the external microphones or their combination is filtered by a peaking filter at a frequency between 1.5 and 2.5 kHz and preferably equal to 2 kHz.

[0044] If the transparent mode is chosen, the wind reduction filter 55 implements a filter in which the signal from at least one of the external microphones or their combination is filtered both by a low shelf filter at a frequency between 200 and 300 Hz and preferably equal to 250 Hz and a peaking filter at a frequency between 800 Hz and 1 kHz and preferably equal to 900 Hz.

[0045] The output of the wind noise reduction filter 55 is connected to a switch 57 suitable for addressing the signal into a noise reduction filter 60 or into a transparency filter 65. The switch 57 is connected to the selector 56 for its control.

[0046] The noise reduction filter 60 is a feed-forward active noise reduction filter based on an estimation of the inverse of the secondary path to cancel the residual noise transmitted through the structure of the earpiece. Such a filter is known per se.

[0047] Similarly, the transparency filter is a predictive active noise amplification filter based on an estimation of the primary path to cancel the damping of the sound by the structure of the earpiece. Such a filter is known per se.

[0048] The outputs of filters 60 and 65 are connected to a summer 70 whose output is connected to a summer 75. Furthermore, the summer 75 is connected to a feedback anti-noise filter 80 directly powered by the internal microphone 45. Alternatively, the filter 80 and the microphone 45 are omitted.

[0049] The output of the adder 75 forms an output 90 of the external noise processing module 25 which is connected to an amplifier, not shown, for exciting the transducer 20.

[0050] Advantageously, an input for a musical signal is also connected to the input of the amplifier.

[0051] [Fig.3] shows the detail of the wind detector 40.

[0052] According to [Fig. 3], the elements of the earpiece illustrated in [Fig. 1] or 2 have the same reference number. Thus, we find the two external microphones 30 and 35.

[0053] The two microphones 30 and 35 are each connected to a module for calculating a fast Fourier transform 95 and 100.

[0054] The signals measured by the microphones 30 and 35 are sampled at a predetermined frequency and are denoted ] for the microphone 30 and y [n] for the microphone 35 and are stored in a buffer memory, n representing the index of the sampling window.

[0055] The fast Fourier transforms of the signals x[n] and y[n] are denoted: n], r [n].

[0056] X[n] and K[n] are vectors with L complex coefficients.

[0057] As wind noise is mainly present at low frequencies, the vectors X[n] and y[n] from the Fourier transforms are truncated in order to keep only the low frequency components and thus reduce the processing load.

[0058] The wind detector 40 comprises two autocorrelation calculation modules 105, 110 for the signals X [ n ] and Y [ n ].

[0059] Each of the autocorrelations XX[n] and yy[n] is calculated recursively from the following calculation:

[0060] = a • XX[n -1] + (1-a) • X[ / î] • conj(X[n])

[0061] yy[«] = a ■ yy[ni] + (l-«) • y[n] • œn / (y[n])

[0062] Where:

[0063] denotes the term-by-term product of the two vectors when between two vectors;

[0064] a is a smoothing parameter between 0 and 1,

[0065] n- 1 is the index of the previous sampling window

[0066] co« / (X[n]), conj(y[n]) is the conjugate of X[n], y[ / i] respectively, that is to say the vector X[ n], respectively Y\n\, whose coefficients are each the conjugates.

[0067] It can be noted here that XX and YY are real numbers because they come from the multiplication of a complex number and its conjugate.

[0068] It can be noted here that XX[n] and yy[n] are real numbers because they come from of the multiplication of a complex number and its conjugate.

[0069] Similarly, a module 115 for calculating the cross-correlation of the signals coming from the microphones 30 and 35 is provided in the wind detector 40. It receives the outputs from the modules 95, 100 for calculating the fast Fourier transforms and ensures the calculation of the cross-correlation noted XY [ / z], recursively according to the following formula:

[0070] xy[n] - a • Xypz-1 ]+ (1-a) • y[ / i] ■ conj(X[n])

[0071] With 0 < a < 1

[0072] The autocorrelations XX[h], and X} [ / ?] and the cross-correlation APp / ] are vectors of complex numbers.

[0073] A coherence calculation module 120 is connected to the output of the two autocorrelation calculation modules 105 and 110 as well as to the output of the cross-correlation calculation module 115.

[0074] The calculation of the coherence noted C[n] between the signals A[h] and y[n] is given by:

[0075] = |xy[ / î]| / ^xx[w]-^]' where denotes the square root term by term

[0076] Module 120 is suitable for calculating the noted average of the square of the consistency C[ n ] by the formula Cmean [zz ] = C^ / z] denotes the arithmetic mean of the coefficients of the vector. Cmean [ n ] is a real number.

[0077] An energy calculation module 125 is connected to the output of the first microphone 30. The energy denoted EilFF [ H ] of the signal x [ n ] from the microphone is calculated using the following calculation:

[0078] EnFFfn] — where denotes the modulus of the complex number jpi]

[0079] The output of the cross-correlation calculation module 115 is connected to a phase variation calculation block 130 A. <p[zz].

[0080] For this purpose, for example, the estimator 130 is suitable for calculating the phase noted çf / z] which is the average of the arguments of the complex coefficients of the cross-correlation vector XI7 [ n ]. (p [ n ] is a real number between 0 and 2ir.

[0081] The module 130 is furthermore capable of calculating the phase difference of the cross-correlation signals XY [n] and XY [?z - 1] corresponding respectively to the current window n and to the previous window n-1:

[0082] A (p [n] - |^[?z] 1]| / 2æ-

[0083] A continuous wind estimation block 135 is connected to the output of the energy calculation module 125, to the output of the coherence calculation module 120 and to the output of the sentence variation calculation module 130.

[0084] The block is suitable for calculating a continuous wind estimator [lUEfn] for each window. WE [n] =1- (1-

[0085] Where al, a2 and a3 are weights.

[0086] WE [ n ] is between 0 and 1.

[0087] A logical wind estimator 140 is connected to the output of the continuous wind estimator 135. The logical estimator 140 is capable of providing at output 145, a logical indicator LE which can take 2 values:

[0088] 1 in the presence of wind, and

[0089] 0 in the absence of wind.

[0090] The value of the logic indicator LE is determined by comparing the value of the continuous wind estimator WE to a predetermined threshold Tl provided by a memory 150.

[0091] To avoid excessively rapid fluctuations, the value of the logic indicator, LE, is only modified if the continuous wind estimator has remained for at least a predetermined duration. preferably finished adjustable and for example equal to one second beyond the predetermined threshold.

[0092] The operation of the algorithm will now be briefly explained.

[0093] The earpiece measures two signals %[ n ] and y [ n ] using two external microphones 30 and 35. The fast Fourier transforms of these two signals are calculated to form vectors with L complex coefficients. Two autocorrelations XX[n] and yy[«], as well as a cross-correlation XY[rt] are calculated recursively.

[0094] A continuous wind presence estimator ILEpz] is calculated from the energy of the signal of one of the two microphones, the coherence calculated from the two autocorrelations and the cross-correlation. The phase variation A (p[n ] is calculated from the cross-correlation XY [ n ] ■ The wind presence estimator WE [ n ] is then established by comparison with the threshold TL

[0095] In the absence of wind, no filter is implemented by the adaptive filter 55.

[0096] In the presence of wind, and depending on the operating mode selected, one of the filters that can be implemented by the adaptive filter 55 is implemented.

Claims

Claims

1. Method for detecting wind implemented in a sound reproduction device and comprising the following steps: - measuring a first sampled external signal (X[ n] ) from a first external microphone (30), - measuring a second sampled external signal from a second external microphone (35), and - calculating a cross-correlation (XF[nJ) between the first and second external signals (Xpi ], F[n]), characterized in that it further comprises: - calculating a variation ( A ç^w]) over the time of a sentence ( tp [n ] ) of the cross-correlation (XY [ n ] ), and - determining the presence of wind as a function at least of the calculated phase variation ( A <p[ / z]).

2. Method according to claim 1, comprising calculating an energy (EnFF[n]) of at least one external signal and determining the presence of wind as a function at least of the calculated phase variation (A ç^ / î]) and the calculated energy (EnFF\n]).

3. Method according to claim 1 or 2, comprising the calculation of an autocorrelation (XX [ / 1], yyp / ]) of each first and second sampled external signal (X [ n ], ÏT^]), a calculation of coherence (£ between the two calculated autocorrelations and the determination of the presence of wind as a function at least of the calculated phase variation (A ç^ / î]) and of the calculated coherence ((7

4. Method according to claims 2 and 3, comprising a step of calculating a wind estimator (WE[n]) equal to a phase variation estimator dependent on the calculated phase variation (A çfrz]), weighted by the product of a coherence estimator dependent on the calculated coherence (r [ / / Tf of an energy estimator dependent on the calculated energy (EnFF[n]) and the phase variation estimator, and in which the determination of the presence of wind comprises the comparison of the wind estimator (VEEfn]) with at least one predetermined threshold (Tl).

5. A method according to claim 4, wherein the wind estimator WE [n] is given by the following formula: W] = 1- {l-( [»] )} - ( 1 - A <p[n] )"’ OÙ Cmea,jn] est la cohérence calculée ; A [ n ] est la variation de phase ; E / lFF[n] est l’énergie d’au moins l’un des premier et second signaux ; et al, a2, et a3 sont des poids.

6. A method according to any preceding claim, comprising calculating the cross-correlation (Xy[ / î]) recursively such that the cross-correlation at the next step is calculated based on the signals sampled at the next step (X[n], / [ / ?]) and the cross-correlation (XK[n - 1]) at the previous step.

7. The method of claim 6, wherein the calculation of the cross-correlation XF [«] is given by the formula: Xy[n]-a • Xy[n-1] + ( 1-a) • F[ n] ■ conj (X[n]) Where: a is a parameter between 0 and 1, X[«] and y [n] are Fourier transforms of the signals measured from the first and second microphones (30, 35), conj (X[n] ) is the conjugate of X[n ] and Xy[n- 1] is the cross-correlation of the two signals at the previous step.

8. A method of processing external noise comprising detecting wind by implementing a method according to any one of the preceding claims, and applying external noise processing dependent on the wind detection.

9. A method of processing external noise according to claim 8, wherein the external noise processing comprises a noise reduction mode (ANC) in which, in the presence of wind, the signal from at least one of the external microphones is filtered by a peak filter at a frequency between 1.5 and 2.5 kHz.

10. A method of processing external noise according to claim 8 or 9, wherein the external noise processing comprises a transparent mode in which, in the presence of wind, the signal from at least one of the external microphones is filtered by both a stepped low pass at a frequency between 200 and 300 Hz and a peak filter at a frequency between 800 Hz and 1 kHz.

11. Wind detection device comprising a first and a second mi- external microphones (30, 35) and processing means suitable for implementing the method according to any one of claims 1 to 7.

12. Sound reproduction device comprising a first and a second external microphone (30, 35) and an external noise processing module (25) capable of implementing a method according to any one of claims 8 to 10.

Images