METHOD AND SYSTEM FOR PROCESSING AN AUDIO SIGNAL INCLUDING AMBISONIC ENCODING

Abstract

A method for processing a sound signal including synchronously acquiring an input sound signal Sinput by means of at least two omnidirectional microphones, encoding the input sound signal Sentréeinput in a sound data D format of the ambisonics type of order R, R being a natural number greater than or equal to one, the encoding step including a directivity optimisation sub-step carried out by means of filters of the Finite Impulse Response filter type. Each of the signals acquired by the microphones is filtered during the directivity optimisation sub-step by a FIR filter, then subtracted from an unfiltered version of each of the other signals in order to obtain N enhanced signals. The present invention also relates to a system for processing the sound signal.

Description

BACKGROUND

1. Field

The present disclosure relates to the field of processing sound signals.

More particularly, the present disclosure relates to the field of recording a 360° sound signal.

2. Brief Description of Related Developments

Methods and systems are known in the prior art for broadcasting 360° video signals. There is a need in the prior art to be able to combine sound signals with these 360° video signals.

Until now, 3D audio has been reserved for sound professionals and researchers. The purpose of this technology is to acquire as much spatial information as possible during the recording to then deliver this to the listener and provide a feeling of immersion in the audio scene.

In the video sector, interest is growing for videos filmed at 360° and reproduced using a virtual reality headset for full immersion in the image: the user can turn his/her head and explore the surrounding visual scene. In order to obtain the same level of precision in the sound sector, the most compact solution involves the use of an array of microphones, for example the Eigenmike by mh acoustics, the Soundfield by TSL Products, and the TetraMic by Core Sound. The polyhedral shape of the microphone arrays allows for the use of simple formulae to convert the signals from the microphones into an ambisonics format. The ambisonics format is a group of audio channels resulting from directional encoding of the acoustic field, and contains all of the information required for the spatial reproduction of the sound field. Equipped with between four and thirty-two microphones, these products are expensive and thus reserved for professional use.

Recent research has focused on encoding in ambisonics format on the basis of a reduced number of omnidirectional microphones. The use of a reduced number of this type of microphones allows costs to be reduced.

By way of example, the publication entitled “A triple microphonic array for surround sound recording” by Rilin CHEN ET AL. discloses an array comprised of two omnidirectional microphones which directivity patterns are virtually modified by applying a delay to one of the signals acquired by the microphones. The resulting signals are then combined to obtain the sound signal in ambisonics format.

One drawback of the method described in this prior art is that the microphones array is placed in a free field. In practice, when an obstacle is placed between the two microphones, diffraction phenomena cause attenuations and phase shifts of the incident wave differentiated according to the frequencies. As a result, the application of a delay to the signal received by one of the microphones will not allows for a faithful reproduction of the sound signal received because the delay applied will be the same at all frequencies.

SUMMARY

The disclosure aims to overcome the drawbacks of the prior art by proposing a method for processing a sound signal allowing the sound signal to be encoded in ambisonics format on the basis of signals acquired by at least two omnidirectional microphones.

The disclosure relates to a sound signal processing method, comprising the steps of:

• • synchronously acquiring an input sound signal S_input by means of N omnidirectional microphones, N being a natural number greater than or equal to two;
  • encoding said input sound signal S_input in a sound data D format of the ambisonics type of order R, R being a natural number greater than or equal to one, said encoding step comprising a directivity optimisation sub-step carried out by means of filters of the Finite Impulse Response FIR filter type, and said encoding step comprising a sub-step of creating an output sound signal S_output in the ambisonics format from enhanced signals derived from the directivity optimisation sub-step;
  • rendering the output sound signal S_output by means of digitally processing said sound data D;

According to the disclosure, during the directivity optimisation sub-step, it is subtracted from each of the signals acquired by the microphones the signals acquired by the N−1 other microphones, each filtered by a FIR filter, in order to obtain N enhanced signals.

In one aspect of the disclosure, the N omnidirectional microphones are integrated into a device.

In one aspect of the disclosure, the FIR filter applied during the directivity optimisation sub-step to each acquired signal is equal to the ratio of the Z-transform of the impulse response of the microphone associated with the signal object of the subtraction over the Z-transform of the impulse response of the microphone associated with the signal to be filtered then subtracted, for an angle of incidence associated with a direction to be deleted.

In one aspect of the disclosure, said microphones are disposed in a circle on a plane, spaced apart by an angle equal to 360°/N.

In one aspect of the disclosure, the method implements four microphones spaced apart by an angle of 90° to the horizontal.

In one aspect of the disclosure, the device is a smartphone and the method implements two microphones, each placed on one lateral edge of said smartphone.

In one aspect of the disclosure, at least one Infinite Impulse Response IIR filter is applied to each of the enhanced signals during the directivity optimisation sub-step in order to correct the artefacts produced by the filtering operations using FIR filters.

In one aspect of the disclosure, the at least one IIR filter is a “peak” type filter, of which a central frequency fc, a quality factor Q and a gain G_dB in decibels can be configured to compensate for the artefacts.

In one aspect of the disclosure, the order R of the ambisonics type format is equal to one.

In one aspect of the disclosure, the creation of the output signal in the ambisonics format is carried out by algebraic operations performed on the enhanced signals derived from the directivity optimisation sub-step in order to create the different channels of said ambisonics format.

The disclosure further relates to a sound signal processing system for implementing the method according to the disclosure. The system according to the disclosure includes means for:

• • acquiring, in a synchronous manner, an input sound signal S_input by means of N microphones, N being a natural number greater than or equal to two;
  • encoding said input sound signal in a sound data D format of the ambisonics type of order R, R being a natural number greater than or equal to one;
  • rendering an output sound signal by means of a digital processing of said sound data D.

According to the disclosure, the sound signal processing system includes means comprising Finite Impulse Response filters for filtering each of the signals acquired by the microphones and subtracting them from each of the other unfiltered original signals in order to obtain N enhanced signals.

BRIEF DESCRIPTION OF THE DRAWINGS

The disclosure will be better understood from the following description and the accompanying figures. These are intended for purposes of illustration only and are not intended to limit the scope of the disclosure.

FIG. 1 shows the different steps of the method according to the disclosure.

FIG. 2 shows a smartphone equipped with two microphones acquiring an acoustic wave.

FIG. 3 shows a block diagram of the sub-steps of optimising the directivity of the microphones and of creating the ambisonics format.

FIG. 4 shows a block diagram for determining Infinite Impulse Response filters used during the directivity optimisation sub-step.

FIG. 5 shows a device including two pairs of microphones, the two directions defined by the two pairs of microphones being orthogonal.

FIG. 6 shows a block diagram for the optimisation of the Left channel in the aspect of the disclosure shown in FIG. 5 comprising four microphones.

FIG. 7 shows a block diagram for the creation of the ambisonics format in the aspect of the disclosure shown in FIG. 5.

FIG. 8 shows two pairs of microphones acquiring an acoustic wave, the two directions defined by the two pairs of microphones forming an angle of strictly less than 90°.

DETAILED DESCRIPTION

With reference to FIG. 1, the present disclosure relates to a method 100 for processing a sound signal, comprising the following steps of:

• • synchronously acquiring 110 an input sound signal S_input by means of N microphones, N being a natural number greater than or equal to two;
  • encoding 120 said input sound signal S_input in a sound data D format of the ambisonics type of order R, R being a natural number greater than or equal to one;
  • rendering 130 an output sound signal S_output by means of digital processing of said sound data D.

In the aspect of the disclosure described hereafter, the acquisition 110 is carried out with a number N of microphones equal to two, and the order R is equal to 1 (the ambisonics format is thus referred to as “B-format”). The channels of the B-format will be denoted in the description below by (W; X; Y; Z) according to usual practice, these channels respectively representing:

• • the omnidirectional sound component (W);
  • the Front-Back sound component (X);
  • the Left-Right sound component (Y);
  • the Up-Down sound component (Z).

Acquisition 110 consists of a recording of the sound signal S_input. With reference to FIG. 2, two omnidirectional microphones M₁, M₂, disposed at the periphery of a device 1, acquire an acoustic wave 2 of incidence θ relative to a straight line passing through the said microphones.

In the shown aspect of the disclosure, the device 1 is a smartphone.

The two microphones M₁; M₂ are considered herein to be disposed along the Y dimension. The reasonings that follow could be conducted in an equivalent manner while considering the two microphones to be disposed along the X dimension (Front-Back) or along the Z dimension (Up-Down), the disclosure not being limited by this choice.

At the end of the acquisition step 110, two sampled digital signals are obtained. y_g is used to denote the signal associated with the “Left channel” and recorded by the microphone M₁ and y_d is used to denote the signal associated with the “Right channel” and recorded by the microphone M₂, said signals y_g, y_d constituting the input signal S_input.

$S_{\mathrm{entrée}}=\left(\begin{array}{c}{y}_g\\ y_d\end{array}\right)S_{\mathrm{input}}$

As shown in FIG. 2, the microphone M₁ first acquires the acoustic wave 2 originating from the left. The microphone M₂ acquires it with a delay relative to the microphone M₁. The delay is in particular the result of:

• • a distance d between the two microphones;
  • the presence of an obstacle, in this case the device 1, causing in particular reflection and diffraction phenomena.

When the acoustic wave 2 has a plurality of frequencies, the delay with which the microphone M₂ acquires said acoustic wave depends on the frequency, in particular as a result of the presence of the device 1 between the microphones causing a diffraction phenomenon.

Similarly, each frequency of the acoustic wave is attenuated in a different manner, as a result of the presence of the device 1 on the one hand, and on the other hand as a function of the directivity properties of the microphones M₁, M₂ dependent on the frequency.

Moreover, since the microphones are both omnidirectional, they both reproduce the entire sound space.

Thereafter, the microphones M₁ and M₂ are sought to be differentiated by virtually modifying their directivity by processing the digital signals recorded, so as to be able to combine the modified signals to create the ambisonics format.

FIG. 3 shows the processing operations applied to the digital signals obtained during the acquisition step 110, within the scope of the encoding step 120 of the method according to the disclosure.

In a directivity optimisation sub-step 121, a filter F₂₁(Z) is applied to the signal y_g of the “Left channel”. The filtered signal is then subtracted from the signal y_d of the “Right channel” by means of a subtractor.

According to the disclosure, the filter F₂₁(Z) is of the Finite Impulse Response (FIR) filter type. Such a FIR filter allows each of the frequencies to be handled independently, by modifying the amplitude and the phase of the input signal over each of the frequencies, and thus allows the effects resulting from the presence of the device 1 between the microphones to be compensated.

By denoting as H₁(Z, θ) and H₂(Z, θ) the respective Z-transforms of the impulse responses of the microphones M₁ and M₂ when integrated into the device 1, in the direction of incidence given by the angle of incidence θ, the filter F₂₁(Z) is determined by the relation:

$F_{21}\left(Z\right)=\frac{H_2\left(Z,\theta =0^{°}\right)}{H_1\left(Z,\theta =0^{°}\right)}$

The choice of a zero angle of incidence θ when determining the filter F₂₁(Z) allows the sound component originating from the left to be isolated. Thus, after subtracting the signals, an enhanced signal y_d* associated with the “Right channel”, from which the sound component originating from the left has been substantially deleted, is obtained.

The directivity of the microphone M₂ is thus virtually modified so as to essentially acquire the sounds originating from the right.

The same operation is carried out in a similar manner for the Left channel. Similarly, a filter F₁₂(Z) is applied to the signal y_d of the Right channel. The filtered signal is then subtracted from the signal y_g of the “Left channel” by means of a subtractor. The filter F₁₂(Z) is a FIR filter defined by the relation:

$F_{12}\left(Z\right)=\frac{H_1\left(Z,\theta ={180}^{°}\right)}{H_2\left(Z,\theta ={180}^{°}\right)}$

The choice of an angle of incidence θ equal to 180° when determining the filter F₁₂(Z) allows the sound component originating from the right to be isolated. Thus, after subtracting the signals, an enhanced signal y_g* associated with the “Left channel”, from which the sound component originating from the right has been substantially deleted, is obtained.

The directivity of the microphone M₁ is thus virtually modified so as to essentially acquire the sounds originating from the left.

In practice, the filters F₂₁(Z) and F₁₂(Z) have properties of high-pass filters and their application produces artefacts. In particular, the frequency spectrum of the enhanced signals y_g*, y_d* is attenuated in the low frequencies and altered in the high frequencies.

In order to correct these defects, at least one filter G₁(Z), G₂(Z) of the Infinite Impulse Response (IIR) filter type is applied to the enhanced signals y_g* and y_d* respectively.

In order to determine the at least one filter G₁(Z) G₂(Z) to be applied, a white noise B is filtered by the filters F₂₁(Z), F₁₂(Z) previously determined, as shown in FIG. 4. The filtered signals are then subtracted from the original white noise B. The comparison of the profiles P, P′ of the output signals with the white noise B allows to determine the one or more filters G₁(Z), G₂(Z) to be applied to correct the alterations of the frequency spectrum as a result of the processing of the signals, during the sub-step 121.

In one aspect of the disclosure, the IIR filters are “peak” type filters, of which a central frequency fc, a quality factor Q and a gain G_dB in decibels can be configured to correct the artefacts. Thus, an attenuated frequency could be corrected by a positive gain, an accentuated frequency could be corrected by a negative gain.

Thus, after filtering by the at least one IIR filter G₁(Z), G₂(Z), a corrected signal Y_G is obtained, representative of the sounds originating from the left and a corrected signal Y_D is obtained, representative of the sounds originating from the right.

Thereafter, with reference to FIG. 3, the output in ambisonics format is created 122.

In order to obtain the omnidirectional component W of the sound signal, the corrected signals Y_D, Y_G are added and the result is normalised by multiplying by a gain K_W equal to 0.5:

$W=\frac{Y_G+Y_D}{2}$

On the basis of the convention according to which the Y component is positive if the sound essentially originates from the left, the Left-Right sound component is obtained by subtracting the corrected signal Y_D associated with the “Right channel” from the corrected signal Y_G associated with the “Left channel”. The result is normalised by multiplying by a factor K_Y equal to 0.5:

$Y=\frac{Y_G-Y_D}{2}$

Given that no information is known on the Front-Back and Up-Down components, the X and Z components are set to zero.

At the end of the encoding step 120, data D in B-format is obtained (in the present aspect of the disclosure, the signals W and Y, the other signals X and Z being set to zero):

$D=\left(\begin{array}{c}\frac{Y_G+Y_D}{2}\\ 0\\ \frac{Y_G-Y_D}{2}\\ 0\end{array}\right)$

The corrected signals Y_G, Y_D of the Left and Right channels respectively can be reproduced by adding and subtracting the signals W and Y:

$\left(\begin{array}{c}{Y}_G\\ Y_D\end{array}\right)=\left(\begin{array}{c}W+Y\\ W-Y\end{array}\right)$

The rendering step 130 consists of rendering the sound signal, thanks to a transformation of the data in ambisonics format into binaural channels.

In one method of implementing the disclosure, the data D in ambisonics format is transformed into data in binaural format.

The disclosure is not limited to the aspect of the disclosure described hereinabove. In particular, the number of microphones used can be greater than two.

In one alternative aspect of the disclosure of the method 100 according to the disclosure, four omnidirectional microphones M₁, M₂, M₃, M₄ disposed at the periphery of a device 1, acquire an acoustic wave 2 of incidence θ relative to a straight line passing through the microphones M₁ and M₂, as shown in FIG. 5.

The two microphones M₁; M₂ are considered herein to be disposed along the Y dimension and the two microphones M₃, M₄ are considered herein to be disposed along the X dimension. The four microphones are disposed in a circle, shown by dash-dot lines in FIG. 5.

At the end of the acquisition step 110, four sampled digital signals are obtained. The following denotations are applied:

• • y_g denotes the signal associated with the “Left channel” and recorded by the microphone M₁;
  • y_d denotes the signal associated with the “Right channel” and recorded by the microphone M₂;
  • x_av denotes the signal associated with the “Front channel” and recorded by the microphone M₃;
  • x_ar denotes the signal associated with the “Back channel” and recorded by the microphone M₄;the said signals y_g, y_d, x_av, X_ar constituting the input signal S_input:

$S_{\mathrm{entrée}}=\left(\begin{array}{c}{y}_g\\ y_d\\ x_{\mathrm{av}}\\ x_{\mathrm{ar}}\end{array}\right)S_{\mathrm{input}}$

With reference to FIG. 6, the directivity optimisation sub-step 121 is shown for this aspect of the disclosure. For clarity purposes, only the processing of the signal y_g associated with the Left channel is shown.

In this aspect of the disclosure, the enhanced signal y_g* is obtained by subtracting the signals y_d, X_av and X_ar respectively filtered by FIR filters F₁₂(Z), F₁₃(Z) and F₁₄(Z) from the signal y_g acquired by the microphone M₁, which filters are defined by:

$F_{12}\left(Z\right)=\frac{H_1\left(Z,\theta ={180}^{°}\right)}{H_2\left(Z,\theta ={180}^{°}\right)}$ $F_{13}\left(Z\right)=\frac{H_1\left(Z,\theta ={90}^{°}\right)}{H_3\left(Z,\theta ={90}^{°}\right)}$ $F_{14}\left(Z\right)=\frac{H_1\left(Z,\theta ={270}^{°}\right)}{H_4\left(Z,\theta ={270}^{°}\right)}$

where H₁(Z, θ), H₂(Z, θ), H₃(Z, θ), H₄(Z, θ) denote the respective Z-transforms of the impulse responses of the microphones M₁, M₂, M₃, M₄ when integrated into the device 1, for an angle of incidence θ.

The choice of the angles of incidence 180°, 90°, 270° when determining the filters allows the sound components respectively originating from the right, from the front and from the back to be isolated.

Thus, after subtracting the signals, an enhanced signal y_g* associated with the “Left channel” is obtained, from which the sound components originating from the right, from the front and from the back have been substantially deleted.

A filter G₃(Z) of the IIR type is then applied to correct the artefacts generated by the filtering operations using FIR filters.

At the end of this step, the corrected signal Y_G is obtained.

Similar processing operations can be applied to the signals of the Right, Front and Back channels, in order to respectively obtain the corrected signals Y_D, X_AV, X_AR.

FIG. 7 describes the sub-step 122 of creating the ambisonics format in the aspect of the disclosure using four microphones described hereinabove.

In order to obtain the omnidirectional component W of the sound signal, the corrected signals Y_D, Y_G, X_AV, X_AR are added and the result is normalised by multiplying by a gain K_W equal to one quarter:

$W=\frac{Y_G+Y_D+X_{\mathrm{AV}}+X_{\mathrm{AR}}}{4}$

On the basis of the convention according to which the Y component is positive if the sound essentially originates from the left, the Left-Right sound component is obtained by subtracting the corrected signal Y_D associated with the “Right channel” from the corrected signal Y_G associated with the “Left channel”. The result is normalised by multiplying by the factor K_Y equal to one half:

$Y=\frac{Y_G-Y_D}{2}$

On the basis of the convention according to which the X component is positive if the sound essentially originates from the front, the Front-Back sound component is obtained by subtracting the corrected signal X_AR associated with the Back channel from the corrected signal X_Av associated with the Front channel. The result is normalised by multiplying by the factor K_x equal to one half:

$X=\frac{X_{\mathrm{AV}}-X_{\mathrm{AR}}}{2}$

In one alternative aspect, the disclosure includes six microphones in order to integrate the Z component of the ambisonics format.

In alternative aspects of the disclosure, the order R of the ambisonics format is greater than or equal to 2, and the number of microphones is adapted so as to integrate all of the components of the ambisonics format. For example, for an order R equal to two, eighteen microphones are implemented in order to form the nine components of the corresponding ambisonic format.

The FIR filters applied to the signals acquired are adapted accordingly, in particular the angle of incidence θ considered for each filter is adapted so as to remove, from each of the signals, the sound components originating from unwanted directions in space.

For example, with reference to FIG. 7, an angle cp between a direction Y through which the microphones M₁ and M₂ pass and a direction X′ through which the microphones M₃ and M₄ pass is strictly less than 90°.

In this aspect of the disclosure, the filter applied to the signal recorded by M₃ and subtracted from the signal acquired by M₁ is given by:

$F_{13}\left(Z\right)=\frac{H_1\left(Z,\theta =\varphi \right)}{H_3\left(Z,\theta =\varphi \right)}$

In this manner, after subtracting the filtered signal from the signal acquired by M₁, an enhanced signal is obtained from which the sound component in the X′ direction has been deleted.

Thus, an ambisonics format of an order greater than or equal to two can be created by adding, for example, microphones in the directions such that φ=45°, φ=90° or φ=135°.

The present disclosure further relates to a sound signal processing system, comprising means for:

• • acquiring, in a synchronous manner, an input sound signal S_input by means of N microphones, N being a natural number greater than or equal to two;
  • encoding the said input sound signal S_input in a sound data D format of the ambisonics type of order R, R being a natural number greater than or equal to one, said means being implemented using filters of the FIR type and using IIR filters of the “peak” type;
  • rendering an output sound signal S_output by means of a digital processing of said sound data D.

This sound signal processing system comprises at least one computation unit and one memory unit.

The above description of the disclosure is provided for the purposes of illustration only. It does not limit the scope of the disclosure.

Claims

1. A method for processing a sound signal the method comprising: synchronously acquiring an input sound signal by means of N omnidirectional microphones, N being a natural number greater than or equal to two;encoding said input sound signal in a sound data D format of the ambisonics type of order R, R being a natural number greater than or equal to one, said encoding step comprising a directivity optimisation sub-step carried out by means of filters of the Finite Impulse Response filter type, and said encoding step comprising a sub-step of creating an output sound signal in the ambisonics format from enhanced signals derived from the directivity optimisation sub-step;rendering the output sound signal by means of a digital processing of said sound data; and

2. The method according to claim 1, wherein the N omnidirectional microphones are integrated into a device.

3. The method according to claim 1, wherein the FIR filter applied during the directivity optimisation sub-step to each acquired signal is equal to the ratio of the Z-transform of the impulse response of the microphone associated with the signal object of the subtraction over the Z-transform of the impulse response of the microphone associated with the signal to be filtered then subtracted, for an angle of incidence associated with a direction to be deleted.

4. The method according to claim 1, wherein the microphones are disposed in a circle on a plane, spaced apart by an angle equal to 360°/N.

5. The method according to claim 4, wherein the method implements four microphones spaced apart by an angle of 90° to the horizontal.

6. The method according to claim 2, wherein the device is a smartphone and wherein the method implements two microphones, each placed on one lateral edge of said smartphone.

7. The method according to claim 1, wherein at least one Infinite Impulse Response (IIR) filter is applied to each of the enhanced signals during the directivity optimisation sub-step in order to correct the artefacts produced by the filtering operations using FIR filters.

8. The method according to claim 7, wherein the at least one IIR filter is a “peak” type filter, of which a central frequency, a quality factor and a gain in decibels can be configured to compensate for the artefacts.

9. The method according to claim 1, wherein the order R of the ambisonics type format is equal to one.

10. The method according to claim 1, wherein the creation of the output signal in the ambisonics format is carried out by algebraic operations performed on the enhanced signals derived from the directivity optimisation sub-step in order to create the different channels of said ambisonics format.

11. A system for processing a sound signal for implementing the method according to claim 1 comprising: acquiring, in a synchronous manner, an input sound signal by means of N microphones, N being a natural number greater than or equal to two;encoding said input sound signal in a sound data format of the ambisonics type of order R, R being a natural number greater than or equal to one; andrendering an output sound signal by means of a digital processing of said sound data;