Multi-channel audio data processing device for automotive vehicle

Abstract

Multi-channel audio data processing device for a motor vehicle A multi-channel audio data processing device for a motor vehicle comprises a data storage (110) receiving multi-channel audio data comprising rank audio data comprising a first audio channel and a second audio channel, a first rank digital filter and a second rank digital filter for each rank. The first rank digital filter and a second rank digital filter of a given rank are determined from a transformation into a digital filter of rank correction data for each row of loudspeakers. A computer (120) applies each first rank digital filter to all or part of the first audio channel of the rank audio data and each second rank digital filter to all or part of the second audio channel of the rank audio data. Fig.1

Description

Title of the invention: Multi-channel audio data processing device for a motor vehicle

[0001] The invention relates to the field of audio processing for motor vehicles.

[0002] The sound quality of the interior of vehicles is a constantly evolving field. For the past few decades, content has been overwhelmingly stereo and now contains a left and right channel (and sometimes a center channel). As progress has been made, the number of speakers inside the passenger compartment has increased, eventually being replicated in each row of seats in the vehicle. Thus, these channels are generally associated with a given row, i.e., front or rear in a conventional vehicle, but potentially more in a van.

[0003] The improvement in sound quality has been accompanied by an increasing increase in the computing power required to implement the desired audio processing. In general, these processing operations are part of a set of processing operations in the passenger compartment, which also consume computing power. Consequently, in recent years, the search for audio quality has therefore been accompanied by a need to find a compromise in available computing power, with all audio processing operations having to fit within an available computing power envelope.

[0004] The sound staging, that is to say the position of the instruments, as well as the tonality of the instruments are currently corrected in a way that the Applicant considers unsatisfactory: hubbub in the bass, spatial blurring of the voice and the instruments, energy holes in the sound staging, voice divided in two, etc.

[0005] These problems are mainly related to the phase and time problems of car-cabin acoustics, which have not been corrected so far. Correcting these problems for each seat of the car at the same time (in the case of multiple listening solutions) adds an additional challenge that is very difficult to solve.

[0006] There are hardware solutions that will not be detailed, but they are not relevant in the context of the invention, which focuses on a software solution. The Applicant has developed an algorithm called Stage that can address some of the problems mentioned, by modifying the phase of certain frequency bands. However, the Applicant thought it could improve this solution.

[0007] The use of manually tuned infinite impulse response (IIR) allpass filters can also be achieved. However, this type of solution is approximate and has very limited accuracy.

[0008] The invention improves the situation. To this end, it proposes a multichannel audio data processing device for a motor vehicle comprising a data storage arranged to receive multichannel audio data comprising rank audio data associated with a respective row of loudspeakers, the multichannel audio data being associated with at least one row, and the rank audio data associated with a given row of loudspeakers comprising a first audio channel and a second audio channel, a first rank digital filter and a second rank digital filter for each row. The first rank digital filter and the second rank digital filter of a given row are determined from correction data comprising rank correction data for each row of loudspeakers associated with the multichannel audio data,the rank correction data associated with a given row of loudspeakers being derived from the measurement of the phase of two signals resulting from the reproduction of, on the one hand, the first audio channel in the given row of loudspeakers and, on the other hand, the second audio channel in the given row of loudspeakers, each time with one of a plurality of frequencies below a threshold between 1.5 kHz and 5 kHz, these phase measurements being carried out on the one hand at a first location associated with the given row of loudspeakers and serving to determine a first set of rank phase differences in which each element is associated with the phase difference at the first location for the measurements obtained for a respective identical frequency for the first audio channel and the second audio channel,and on the other hand at a second location associated with the given row of loudspeakers and used to determine a second set of row phase differences in which each element is associated with the phase difference at the second location for the measurements obtained for a respective identical frequency for the first audio channel and the second audio channel. Each element of the row correction data being associated with one of the frequencies of the plurality of frequencies below a threshold between 1.5 kHz and 5 kHz, each element Q is associated with one of the frequencies of the plurality of frequencies below a threshold between 1.5 kHz and 5 kHz, according to the formula C; - Dj + Pj + S^Fj where i is the index of a given frequency in the plurality of frequencies below a threshold between 1.5 kHz and 5 kHz, Dt is calculated according to the formula t„„-1 / 52^^52^ j where Rlf is the value of the element, LdI1 \ sinÆlrsin«2; / of index i in the first set of rank phase differences, and RZ; is the value of the element of index i in the second set of rank phase differences, Pi is a phase correction value equal to 0 or ir and a function of the value of the sum cosZ?l7- + cosÆ2f-, S is a distribution factor between -1 and 1, and Fi is calculated according to the formula _ R2- + tt) mod 2tt — TT First rank digital filter and the second rank digital filter are taken from a transformation of the data of rank correction into a finite impulse response filter or an infinite impulse response filter. This device further comprises a calculator arranged to apply each first digital rank filter of a given rank to all or part of the first audio channel of the audio data of rank associated with the given rank and each second digital rank filter of a given rank to all or part of the second audio channel of the audio data of rank associated with the given rank.

[0009] This device is particularly advantageous because it allows the phase alignment between the various frequencies to be corrected for each rank of the vehicle with a contained computing power. Unlike the Stage algorithm or the manual adjustment of the all-pass filters, this solution is based on metrology, and the phase problems in the car are measured and the correction is carried out using a measuring system. In addition, this correction can be customized, and be balanced between the first location and the second location, or be centered on the first location or the second location, certainly to the detriment of the other location.

[0010] According to various embodiments, the invention may have one or more of the following characteristics: - the first rank digital filter and the second rank digital filter are determined from rank correction data in which Pi value is calculated according to the formula TT if Cos\j + Cos2; < h and P^ — 7T where h is a hysteresis factor chosen 0 if Cos^ + Cos2î > - h and PiA = 0 P* Æ if Cadj + Coslj< -h and 0 if Cos^ + Cos2i >h and PiA = tt in the range [0; 0.2], - the calculator is further arranged: * on the one hand receiving as input input rank audio data, to separate them into low frequency rank audio data whose frequencies are lower than said threshold between 1.5 kHz and 5 kHz, and into high frequency audio data whose frequencies are higher than said threshold between 1.5 kHz and 5 kHz, and * on the other hand to return output rank audio data comprising a first audio channel combining a signal taken from the first audio channel of the high frequency rank audio data and a signal taken from the first audio channel of the low frequency rank audio data corrected by the first rank digital filter and a second audio channel combining a signal taken from the second audio channel of the high frequency rank audio data and a signal taken from the second audio channel of the low frequency rank audio data corrected by the second rank digital filter, - the computer is arranged to subsample the low-frequency audio data, on the one hand to apply the first digital filter of rank to the first audio channel of the subsampled low-frequency audio data and oversample the resulting signal to produce the signal taken from the first audio channel of the low-frequency audio data corrected by the first digital filter of rank, and on the other hand, to apply the second digital filter of rank to the second audio channel of the subsampled low-frequency audio data and oversample the resulting signal to produce the signal taken from the second audio channel of the low-frequency audio data corrected by the second digital filter of rank, - the calculator is arranged to produce a signal drawn from the first audio channel of the high-frequency rank audio data and a signal drawn from the second audio channel of the high-frequency rank audio data having a delay chosen as a function of the rank correction data, and - the first rank digital filter and the second rank digital filter are obtained by transformation into a finite impulse response filter according to a distribution of the rank correction data such that C'first audio channel) -and Ci (second audio channel) = -(1- a)*^ where a is a real number between 0 and 1.

[0011] The invention also relates to a method for processing a multichannel audio signal for a motor vehicle comprising the following operations: a) receiving multichannel audio data comprising rank audio data associated with a respective row of loudspeakers, the multichannel audio data being associated with at least one row, and the rank audio data associated with a given row of loudspeakers comprising a first audio channel and a second audio channel, b) obtaining a first rank digital filter and a second rank digital filter for each row, the first rank digital filter and a second rank digital filter of a given row being determined from correction data comprising rank correction data for each row of loudspeakers associated with the multichannel audio data, the rank correction data associated with a given row of loudspeakers being derived from the measurement of the phase of two signals resulting from the reproduction of the first audio channel in the given row of loudspeakers and the second audio channel in the given row of loudspeakers,each time with one of a plurality of frequencies below a threshold between 1.5 kHz and 5 kHz, these phase measurements being carried out on the one hand in a first location associated with the given row of loudspeakers and used to determine a first set of row phase differences in which each element is associated with, the phase difference at the first location for measurements obtained for a respective identical frequency for the first audio channel and the second audio channel, and on the other hand at a second location associated with the given row of loudspeakers and used to determine a second set of row phase differences in which each element is associated with the phase difference at the second location for measurements obtained for a respective identical frequency for the first audio channel and the second audio channel, each element of the row correction data being associated with one of the frequencies of the plurality of frequencies below a threshold between 1.5 kHz and 5 kHz, each element C, is associated with one of the frequencies of the plurality of frequencies below a threshold between 1.5 kHz and 5 kHz, according to the formula C, = D) + Pj + S*F{ where i is the index of a given frequency in the plurality of frequencies below a threshold between 1,5 kHz and 5 kHz, is calculated according to the formula / cosR^-cosRii \ oq r 1; csl |has the value of the element with index i \ sinÆl^siiJK; / , in the first set of rank phase differences, and R^j is the value of the element of index i in the second set of rank phase differences, P, is a phase correction value equal to 0 or ir and a function of the value of the sum cosÆl^ + coslK;, S is a distribution factor between -1 and 1, and Fj is calculated according to the formula (RI-- R2 + tt) mod 2tt — Æ' 'c First rank digital filter and the second rank digital filter being derived from a transformation of the rank correction data into a finite impulse response filter or an infinite impulse response filter, and (c) applying each first rank digital filter of a given rank to all or part of the first audio channel of the rank audio data associated with the given rank and each second rank digital filter of a given rank to all or part of the second audio channel of the rank audio data associated with the given rank.

[0012] According to various embodiments, this method may have one or more of the following characteristics: - the first rank digital filter and the second rank digital filter of operation b) are determined from rank correction data in which the value Pi is calculated according to the formula TT if Cos 1; + Cos2; < h and — TT where h is a hysteresis factor chosen 0 if Cos + Cos2} > - h and PiA = 0 n if Cos+ Cos2; < -h and P^ = 0 0 if Cos lf + Cos2( > h and P^ = tt in the range [0; 0,2], - operation a) comprises separating the rank audio data into low-frequency rank audio data whose frequencies are lower than said threshold between 1.5 kHz and 5 kHz, and into high-frequency audio data whose frequencies are higher than said threshold between 1.5 kHz and 5 kHz, and operation c) comprises returning rank audio data comprising a first audio channel combining a signal taken from the first audio channel of the high-frequency rank audio data and a signal taken from the first audio channel of the low-frequency rank audio data corrected by the first rank digital filter and a second audio channel combining a signal taken from the second audio channel of the high-frequency rank audio data and a signal taken from the second audio channel of the low-frequency rank audio data corrected by the second rank digital filter, - operation c) comprises cl) downsampling the low frequency audio data, on the one hand to apply the first rank digital filter to the first audio channel of the downsampled low frequency audio data and upsample the resulting signal to produce the signal derived from the first audio channel of the low frequency audio data corrected by the first rank digital filter, and on the other hand, to apply the second rank digital filter to the second audio channel of the downsampled low frequency audio data and upsample the resulting signal to produce the signal derived from the second audio channel of the low frequency audio data corrected by the second rank digital filter,and c2) combining on the one hand the signal taken from the first audio channel of the low frequency rank audio data corrected by the first rank digital filter with the first audio channel of the high frequency rank audio signal and on the other hand the signal taken from the second audio channel of the low frequency rank audio data corrected by the second rank digital filter with the second audio channel of the high frequency rank audio signal to produce the returned rank audio data, , - operation c) comprises producing a signal drawn from the first audio channel of the high frequency rank audio data and a signal drawn from the second audio channel of the high frequency rank audio data having a delay selected as a function of the rank correction data, and - the first rank digital filter and the second rank digital filter are obtained by transformation into a finite impulse response filter according to a distribution of the rank correction data such that C'first audio channel) -and Ci (second audio channel) = -(1- a)*^ where a is a real number between 0 and 1.

[0013] The invention also relates to a computer-implemented computer program comprising instructions for executing the method according to the invention, and a data storage medium on which such a computer program is recorded.

[0014] Other characteristics and advantages of the invention will appear more clearly on reading the following description, taken from examples given for illustrative and non-limiting purposes, taken from the drawings in which: - [Fig.l] represents a schematic top view of a vehicle implementing the device according to the invention, and - [Fig.2] represents a generic block diagram of the device used in [Fig.l].

[0015] The drawings and the description below contain, for the most part, elements of a certain character. They may therefore not only serve to better understand the present invention, but also contribute to its definition, if necessary.

[0016] [Fig.l] represents a schematic top view of a vehicle implementing the device according to the invention.

[0017] As can be seen in this figure, a vehicle 2 has two rows of seats: a front row 4 and a rear row 6. The front row 4 includes a driver's seat 8 and a passenger seat 10. The rear row 6 includes a left rear seat 12 and a right rear seat 14. Alternatively, the rear row 6 could have more seats, and the vehicle could have a third or more rows. Still as a variant, there may be a single row of seats.

[0018] In the example described here, the front row 4 comprises a left front speaker 16 and a right front speaker 18, while the rear row 6 comprises a left rear speaker 20 and a right rear speaker 22.

[0019] The Applicant, seeking to improve the Stage algorithm, was interested in carrying out measurements in vehicle passenger compartments. Indeed, obtaining a wide and coherent sound scene at a given location relies largely on the fact that, for a given signal, the frequencies it contains reach this location at the same time. However, the sound processing and the particular shape of the passenger compartment mean that the frequencies do not arrive at the ear in the same way and do not propagate in the same way. This is particularly pronounced in the case of the part of multichannel audio signals which is identical for the channels and is generally designated by the term "mono part" or "correlated part".

[0020] In doing so, the Applicant discovered that it is advantageous to process the phase row by row, because the seats in a given row mainly receive the sound channels from the loudspeakers in that row. Thus, initially, the Applicant carried out separate measurements of linear frequency sweeps ("frequency sweep" in English) at each location in each row, for each audio channel in each row.

[0021] Thus, it was possible to obtain vectors which represent, for a large number of frequencies, the phase of the signal which reaches each location. To be more explicit, these vectors can be designated by designating the location at which they were measured, the rank to which the location belongs and the channel from which they are taken.

[0022] For example, the vector associated with location 1 in row 1 (i.e., pilot seat 8 in [Fig.l]) and the first audio channel (e.g., the left channel) of row 1 will be denoted FL1, the vector associated with location 1 in row 1 (i.e., pilot seat 8 in [Fig.l]) and the second audio channel (e.g., the right channel) of row 1 will be denoted FRI, the vector associated with location 2 in row 1 (i.e., passenger seat 10 in [Fig.l]) and the first audio channel (e.g., the left channel) of row 1 will be denoted FL2, the vector associated with location 2 in row 1 (i.e., passenger seat 10 in [Fig.l]) and the second audio channel (e.g., the right channel) of row 1 will be denoted FR2. For rank 2, the notations will be similar, but the first letter of the vector will change to indicate that rank (e.g. "G", or any other letter).

[0023] The vectors each comprise the phase measurements for a plurality of frequencies below a chosen threshold. The Applicant has discovered that a threshold between 1.5 kHz and 5 kHz gives the best results.

[0024] Indeed, the phase alignment effects are particularly noticeable at the lowest frequencies, and are no longer noticeable beyond 5kHz: feeling the phase effects at this frequency requires almost perfect hearing, while not moving the head more than 5 cm - and the higher the frequency, the more the precision of the human ear decreases while the possible movement of the head decreases. Keeping the head still within a limit of 5 cm in a motor vehicle already seems unlikely enough while driving.

[0025] As will be seen below, the filter which corresponds to the correction calculated using the phase vector measurements can be implemented by a finite impulse response filter ("FIR" in English) or an infinite impulse response filter ("IIR" in English).

[0026] The threshold value determines on the one hand the frequencies for which the phase alignment is carried out, but also by the Nyquist condition the size of the filter which must be used to implement it. However, the higher the frequency, the higher the number of coefficients in the filter must be for comparable correction accuracy, and the greater the computing power required to implement it. The lower the threshold, the less computing power is required, but the correction is then less good. The Applicant has determined that with a threshold below 1.5 kHz, the correction no longer presents an interesting quality. The range between 1.5 kHz and 5 kHz has therefore been determined as offering the best compromise. for the implementation of the invention. Preferably, the range for the threshold may be between 2 kHz and 3 kHz, and the preferred value of the threshold is 3 kHz because it represents the best compromise in computing power / correction quality in the Applicant's tests. As will be seen below, the correction is implemented by means of a digital filter. The Applicant has found that it is preferable for the resolution of the phase measurement for the corrections to be greater than or equal to that of the filter.

[0027] Once the phase vectors were measured, the Applicant worked on a solution to correct the phase alignment at each location of a given rank. Its work showed that it is interesting to work on the phase differences at each location. Thus, the vectors FL1, FRI, FL2 and FR2 were assembled into two phase difference vectors RI and R2 according to the formulas R1=FL1-FR1 and R2=FL2-FR2. This means that the vector RI contains the phase difference, at location 1 of rank 1 between the phases of the first audio channel and the second audio channel. Similarly, the vector R2 contains the phase difference, at location 2 of rank 1 between the phases of the first audio channel and the second audio channel.

[0028] It then becomes possible to correct the alignment of the phases of the various vectors by applying a correction based on three correction factors: (a) a phase inversion factor P, which consists, for each frequency, of adding a phase correction of value Æ if the sum of the cosines of the phase difference at the two locations is less than 0, b) a symmetrization regularization factor D, and c) a correction location adjustment factor S*F.

[0029] The Applicant has discovered that the phase inversion factor P can be determined in two distinct ways: with and without hysteresis. The correction without hysteresis is based on the assumption that the measurements are perfectly reliable. In this case, the value of each element Pi of the phase inversion vector P is determined according to the following formula: - if cosR 1; + cos / ?2(- < 0, P} = tt - if cosKl; + cosR2j > 0, P{ = 0.

[0030] However, the measurement of the phases and therefore of their difference is necessarily noisy. This means that around 0 for the sum cosi?!, + cosR2^ the value Pi of neighboring frequencies can vary greatly, and this only because of the measurement noise, which is likely to generate artifacts in the correction. Even in the case if the measurements were perfect, in the case where the sum cosRl,+ cos / ?2;- varies very quickly around 0, this can pose a problem in the design of a digital filter. Indeed, it is complex, even impossible to successively vary a digital filter so quickly.

[0031] For this reason, the Applicant determined the correction with hysteresis, according to the following formula: TT if cosR 1,-+cosÆ2; < h and P = n I l {“1 0 if cosÆl, + cosR2i> -h et= 0 7i if cosRlz- + cosÆ2; < -h and P^ = 0 0 if cosRïi + cosR2i >h and P^ = æ

[0032] This means that, around 0, and with in a range of width h, the value of Pi does not change in order to preserve the stability of the phase correction. It is therefore necessary that the value of cosUly + cosl^ whose absolute value exceeds h to go from a correction of 0 to or from Æ to 0. Consequently, some frequencies will not be corrected when they should have been, but artifacts due to measurement noise will be avoided. The Applicant has determined that values of h chosen in the range [0; 0.2] give the best results. Incidentally, the value of h equal to zero corresponds to the formula without hysteresis.

[0033] The symmetrization regularization factor D is determined according to the following formula: cosXl,-cosX2,. \ sinX l-sinXÎ, /

[0034] The regularization factor D allows to obtain a symmetry of performance between the left seat and the right seat: the recombination of the right and left channels at the two seats will be identical. Perceptually, this allows for the most similar soundstage quality and frequency response to each other at the different locations in the car.

[0035] Finally, the correction location adjustment factor S*F allows the phase correction to be adjusted in order to favor the correction at a particular location in the rank. This is obtained first by determining the vector F whose elements Fi are calculated by the formula — R^-+ yj mod 2tt tt' and in 'c scalar adjustment S to vector F.

[0036] If S is equal to 0, then the phase correction is perceived as the same at the first location and at the second location. If S is equal to 1, the phase correction will be optimal for location 1. If S is equal to -1, the phase correction will be optimal for location 2. This factor therefore makes it possible to modulate the perceived quality of the phase correction for a given location. Thus, when a person is driving alone, it will be very advantageous to use a value S equal to 1. Similarly, it may be advantageous to switch between various filters corresponding to various values of S depending on the situations: for example, in the event of an interruption of a sound program to give information (for example traffic news on the radio or interruptions of GPS guides), it may be advantageous to use a filter in which S is 1 to facilitate driver understanding.

[0037] Finally, the phase correction vector can be defined as the sum of the three vectors of the phase inversion factor, the symmetrization regularization factor D, and the correction location adjustment factor, i.e. a correction vector C of which each element Ci is obtained by the formula Ct = D{ + P; +

[0038] Once the correction vector C has been determined for each rank and for each desired value of S, it is transformed into a digital filter so that it can be used in the motor vehicle. For this, in the context of a finite impulse response filter, a design based on least squares ("Least mean squares" in English, or "LMS"f) or a design by integration and windowing can be used.

[0039] In the case of an infinite impulse response filter, the design may be based on least squares. Great care must then be taken to manage the cases of divergence which characterize infinite impulse response filters. For this reason, although they consume more computing power, finite impulse response filters are preferred by the Applicant.

[0040] This results in one or more digital filters for each rank and for each value of S that it is desired to use. These filters are to be applied exclusively to the first audio channel, or to the second audio channel.

[0041] Alternatively, the Applicant has discovered that it may be advantageous to produce digital filters for each audio channel, which makes it possible to distribute the correction and therefore limits the risks of saturation ("filler saturation" in English) of the digital filter. In this case, a correction factor a is used to define a correction filter for the first channel and a correction filter for the second channel, according to the formula G (first audio channel) = a*Ci and q (second audio channel) = - (1 - a)*Cj with a chosen in the range between 0 and 1. The value of 0.5 makes it possible to optimize the risk of saturation since half of the correction is carried by each audio channel in the resulting digital filters.

[0042] Once this research work has been carried out, the Applicant was therefore able to implement a multi-channel audio data processing device for a motor vehicle 100 as shown in [Fig.2].

[0043] The device 100 comprises a data storage 110 and a calculator 120.

[0044] The data storage 100 receives two types of data: on the one hand the data multichannel audio to be processed 104, and on the other hand the data defining the digital phase correction filters for each rank 108.

[0045] The data storage 100 can be any type of data storage suitable for receiving digital data: hard disk, flash memory hard disk, flash memory in any form, RAM, optical disk, locally or cloud distributed storage, etc.

[0046] The computer (PC) 120 may include one or more processors (P) 122. In some embodiments in which the computer 120 comprises a programmable processor, a computer program product (CPP) 124 (or "computer program product" in English) may be provided to implement the phase correction processing. The computer program CPP 124 stores a computer program (CP) 126 which includes computer readable instructions (CRI) 128 (or "computer readable instructions" in English). The computer program CPP 124 may be stored on a computer readable medium (CRM) 130 (or "computer readable medium" in English) which may be a non-transitory computer readable medium (or "non-transitory computer readable medium" in English), such as for example a magnetic medium (hard disk, SSD, magnetic tape, etc.), an optical medium (CD disk, DVD disk, Blu-ray disk, etc.), a memory (RAM, flash memory, etc.), distributed or cloud storage, etc. The computer-readable medium (CRM) 130 may also be stored in the data storage 100.

[0047] The processor(s) P 122 may be any processor suitable for the calculations described below. Such a processor may be produced in any known manner, in the form of a microprocessor for a personal computer, laptop, tablet or smartphone, a processor dedicated to signal processing ("DSP" in English), a dedicated chip of the FPGA or SoC type, a computing resource on a grid or in the cloud, a cluster of graphics processing units ("GPUs"), a microcontroller, or any other form suitable for providing the computing power necessary for the implementation described below. One or more of these elements may also be produced in the form of specialized electronic circuits such as an ASIC. A combination of processor(s) and electronic circuits may also be envisaged. Processors dedicated to machine learning may also be envisaged.

[0048] Since the correction filter is limited to frequencies below the threshold between 1.5 kHz and 5 kHz, the computer 120 proceeds by filtering the incoming multi-channel audio signal into a high-frequency signal whose frequencies are above the threshold, and a low-frequency signal whose frequencies are below the threshold. This means that each channel of each rank of the multi-channel audio signal is separated into a low-frequency component and a high-frequency component.

[0049] Then, the calculator 120 applies to each channel of each rank of the low-frequency signal the digital filter which corresponds to it, then it recombines the signals results with the high frequency signal of each channel to produce the corrected audio signal.

[0050] Advantageously, the Applicant has discovered that two treatments can be applied to further improve the correction and / or the computing power consumed: - the calculator 120 can be arranged to downsample the low-frequency signal, apply the digital filters to it, then upsample it before recombining it with the high-frequency signal.

[0051] This allows the use of even less expensive digital filters. - the computer 120 can determine a delay factor for the high-frequency signal as a function of the vector C determined for the rank concerned. Indeed, the vector C will have the effect of delaying certain of the frequencies in the low-frequency signal, and it is desirable to align the high-frequency signal as much as possible with these delays as well as the delay induced by the design of the digital filter.

Claims

1. Claims Multi-channel audio data processing device for a motor vehicle comprising - a data storage (110) arranged to receive * multi-channel audio data comprising rank audio data associated with a respective speaker rank, the multi-channel audio data being associated with at least one rank, and the rank audio data associated with a given speaker rank comprising a first audio channel and a second audio channel, * a first rank digital filter and a second rank digital filter for each rank, the first rank digital filter and a second rank digital filter of a given rank being determined from correction data comprising rank correction data for each row of loudspeakers associated with the multichannel audio data, the rank correction data associated with a given row of loudspeakers being derived from the measurement of the phase of two signals resulting from the reproduction of, on the one hand, the first audio channel in the given row of loudspeakers and, on the other hand, the second audio channel in the given row of loudspeakers, each time with one of a plurality of frequencies below a threshold of between 1.5 kHz and 5 kHz,these phase measurements being carried out on the one hand at a first location associated with the given row of loudspeakers and serving to determine a first set of row phase differences in which each element is associated with the phase difference at the first location for the measurements obtained for a respective frequency identical for the first audio channel and the second audio channel, and on the other hand at a second location associated with the given row of loudspeakers and used to determine a second set of row phase differences in which each element is associated with the phase difference at the second location for the measurements obtained for a respective identical frequency for the first audio channel and the second audio channel, each element of the row correction data being associated with one of the frequencies of the plurality of frequencies below a threshold between 1.5 kHz and 5 kHz, each element Cf is associated with one of the frequencies of the plurality of frequencies below a threshold between 1.5 kHz and 5 kHz, according to the formula Ci = ^t + Pi + S^Fj where * i is the index of a given frequency in the plurality of frequencies below a threshold between 1.5 kHz and 5 kHz, * D{ is calculated according to the formula \ where ri. csl |a \ siuRlfSinÆS; / value of the element of index i in the first set of phase differences of rank, and is the value of the element of index i in the second set of phase differences of rank * Pi is a phase correction value equal to 0 or ir and a function of the value of the sum cosZ?l;+ cos / f2j, * S is a distribution factor between -1 and 1, and * F{ is calculated according to the formula R2- + tt) mod Zît — TV the first rank digital filter and the second rank digital filter being derived from a transformation of the rank correction data into a finite impulse response filter or an infinite impulse response filter, the device (100) further comprising a calculator (120) arranged to apply each first rank digital filter of a given rank to all or part of the first audio channel of the rank audio data associated with the given rank and each second rank digital filter of a given rank to all or part of the second audio channel of the rank audio data associated with the given rank.;

2. The device of claim 1, wherein the first rank digital filter and the second rank digital filter are determined from rank correction data in which the value Pi is calculated according to the formula tt if Cos lr- + Cos2{ < h and P iA-jr 0 if Cos\ + Cos!:> -h and Pn = 0 p. — 1 tt if Cos\i + Coslj < -h and Pi} - 0 0 if Cos 1, + Coslj > h and PiA - tt where h is a hysteresis factor selected from the range [0; 0.2].

3. Device according to one of the preceding claims, in which the calculator (120) is further arranged: * on the one hand to receive as input audio data of input rank, to separate them into audio data of low frequency rank whose frequencies are lower than said threshold between 1.5 kHz and 5 kHz, and into high-frequency audio data whose frequencies are higher than said threshold between 1.5 kHz and 5 kHz, and * on the other hand to return output rank audio data comprising a first audio channel combining a signal taken from the first audio channel of the high-frequency rank audio data and a signal taken from the first audio channel of the low-frequency rank audio data corrected by the first rank digital filter and a second audio channel combining a signal taken from the second audio channel of the high-frequency rank audio data and a signal taken from the second audio channel of the low-frequency rank audio data corrected by the second rank digital filter.

4. Device according to claim 3, in which the computer (120) is arranged to subsample the low frequency audio data, on the one hand to apply the first digital filter of rank to the first audio channel of the subsampled low frequency audio data and oversample the resulting signal to produce the signal drawn from the first audio channel of the low frequency audio data corrected by the first digital filter of rank, and on the other hand, to apply the second digital filter of rank to the second audio channel of the subsampled low frequency audio data and oversample the resulting signal to produce the signal drawn from the second audio channel of the low frequency audio data corrected by the second digital filter of rank.

5. Apparatus according to claim 3 or 4, wherein the computer (120) is arranged to produce a signal drawn from the first audio channel of the high frequency rank audio data and a signal drawn from the second audio channel of the high frequency rank audio data having a delay chosen as a function of the rank correction data.

6. Device according to one of the preceding claims, in which the first digital rank filter and the second digital rank filter are obtained by transformation into a finite impulse response filter according to a distribution of the rank correction data such that first audio channel) =a*C{ and

7. Cj (second audio channel) = - (1 - a)*Q where a is a real number between 0 and 1. A method for processing a multi-channel audio signal for a motor vehicle comprising the following operations: a) receiving multichannel audio data comprising rank audio data associated with a respective row of speakers, the multichannel audio data being associated with at least one row, and the rank audio data associated with a given row of speakers comprising a first audio channel and a second audio channel, b) obtaining a first rank digital filter and a second rank digital filter for each row, the first rank digital filter and a second rank digital filter of a given row being determined from correction data comprising rank correction data for each row of speakers associated with the multichannel audio data, the rank correction data associated with a given row of speakers being derived from the measurement of the phase of two signals resulting from the reproduction of the first audio channel in the given row of speakers and the second audio channel in the given row of speakers,each time with one of a plurality of frequencies below a threshold between 1.5 kHz and 5 kHz, these phase measurements being carried out, on the one hand at a first location associated with the given row of loudspeakers and used to determine a first set of row phase differences in which each element is associated with the phase difference at the first location for the measurements obtained for a respective identical frequency for the first audio channel and the second audio channel, and on the other hand at a second location associated with the given row of loudspeakers and used to determine a second set of row phase differences in which each element is associated with the phase difference at the second location for the measurements obtained for a respective identical frequency for the first audio channel and the second audio channel, each element of the row correction data being associated with one of the frequencies of the plurality of frequencies below a threshold between 1.5 kHz and 5 kHz, each element Q is associated with one of the frequencies of the plurality of frequencies below a threshold between 1.5 kHz and 5 kHz, according to the formula Ci = ^t + Pi + S^Fj where * i is the index of a given frequency in the plurality of frequencies below a threshold between 1.5 kHz and 5 kHz, * D{ is calculated according to the formula \ where ri.csl |a \siuRlfSinÆS; / value of the element with index i in the first set of phase differences of rank, and Ri is the value of the element with index i in the second set of phase differences of rank * Pi is a phase correction value equal to 0 or ir and a function of the value of the sum cosZ?l; + cos / f2j, * S is a distribution factor between -1 and 1, and * F{ is calculated according to the formula R2- + tt) mod 2tt — TV the first rank digital filter and the second rank digital filter being derived from a transformation of the rank correction data into a finite impulse response filter or an infinite impulse response filter, c) applying each first rank digital filter of a given rank to all or part of the first audio channel of the rank audio data associated with the given rank and each second rank digital filter of a given rank to all or part of the second audio channel of the rank audio data associated with the given rank.

8. The method of claim 7, wherein the first rank digital filter and the second rank digital filter of operation b) are determined from rank correction data in which the value Pi is calculated according to the formula tt if Cosl^ + Coslf < h and PiA = tt 0 if Cos L + Cos2f > - h and Pn = 0 P.= { L ît if Coslf + Cos2( £ -h and PiA = 0 0 if Cos I,- + Cos2{ > h and P^ j = tt where h is a hysteresis factor chosen from the range [0; 0,2].

9. Method according to one of claims 7 and 8, in which operation a) comprises separating the audio data of rank into low frequency audio data whose frequencies are lower than said threshold between 1.5 kHz and 5 kHz, and into high frequency audio data whose frequencies are higher than said threshold between 1.5 kHz and 5 kHz, and operation c) comprises returning rank audio data comprising a first audio channel combining a signal from the first audio channel of the high frequency rank audio data and a signal from the first audio channel of the low frequency rank audio data corrected by the first rank digital filter and a second audio channel combining a signal from the second audio channel of the high frequency rank audio data and a signal from the second audio channel of the low frequency rank audio data corrected by the second rank digital filter.

10. The method of claim 9, wherein step c) comprises cl) downsampling the low-frequency audio data, on the one hand to apply the first rank digital filter to the first audio channel of the downsampled low-frequency audio data and upsample the resulting signal to produce the signal derived from the first audio channel of the low-frequency audio data corrected by the first rank digital filter, and on the other hand, to apply the second rank digital filter to the second audio channel of the downsampled low-frequency audio data and upsample the resulting signal to produce the signal derived from the second audio channel of the low-frequency audio data corrected by the second rank digital filter,and c2) combining on the one hand the signal taken from the first audio channel of the low frequency rank audio data corrected by the first rank digital filter with the first audio channel of the high frequency rank audio signal and on the other hand the signal taken from the second audio channel of the low frequency rank audio data corrected by the second rank digital filter with the second audio channel of the high frequency rank audio signal to produce the returned rank audio data.,

11. A method according to claim 9 or 10, wherein step c) comprises producing a signal from the first audio channel of the high frequency rank audio data and a signal from the second audio channel of the high frequency rank audio data having a delay selected as a function of the rank correction data.

12. Method according to one of claims 7 to 11, in which the first rank digital filter and the second rank digital filter are obtained by transformation into a finite impulse response filter according to a distribution of the data of 20 rank correction such that first audio channel) - a*Ct and Ci (second audio channel) = -(1- a)*Ci where a is a real number between 0 and 1.

13. A computer-implemented computer program comprising instructions for executing the method according to one of claims 7 to 12.

14. Data storage medium on which the computer program according to claim 13 is recorded.

Images