Improved MIMO control method, associated control system and computer program

Abstract

The invention relates to a method for controlling an acoustic field which is generated by a sound reproduction system including a plurality of loudspeakers, the method being implemented by a control system which is connected to the loudspeakers and includes a plurality of microphones positioned in the listening zone, the method, implemented by a computer, consisting in: defining (111) first and second frequency bands in the operating range of the sound reproduction system; acquiring (113) a response matrix (H) of the loudspeakers; defining (115) a first target (d(1)) for the first band and a second target (d(2)) for the second band; determining (117) a first filter (g(1)) for the first band on the basis of the response matrix (H) and the first target and a second filter (g(2)) for the second band on the basis of the response matrix (H) and the second target; and combining (119) the first and second filters to obtain the global filter.

Description

[0001]TITLE: Improved MIMO control method; Associated control system and computer program. The present invention relates to methods for controlling an acoustic field produced by a sound reproduction system. Such a system comprises several loudspeakers. A loudspeaker, even if it may comprise more than one transducer, is powered by a single electrical excitation signal generally originating from an amplifier. For example, loudspeakers capable of reproducing all frequencies (bass, medium and treble), loudspeakers capable of reproducing only medium or treble frequencies, and loudspeakers capable of reproducing only low frequencies, or subwoofers, are known.Such a sound reproduction system is implemented in a room such as a mixing studio, a cinema or concert hall, or in the living room of an apartment or a single-family home for home cinema applications. Generally, the reproduction of the sound field is done in a listening area inside the room. However, the walls of the room (and any obstacles inside this room) are the source of reflections that amplify, attenuate or cancel the sound field at certain points in the listening area. These alterations of the sound field, which depend on the frequency, are different depending on the position of the listener in the listening area. To obtain a satisfactory listening experience regardless of the position inside the listening area, it is necessary to control the sound waves emitted by the loudspeakers of the sound reproduction system.Such control depends on a very large number of parameters, such as the number, position and orientation of the loudspeakers, the geometry of the listening room, the materials constituting the walls of this room and their absorption or reflection properties depending on the frequency, the position of any obstacles, etc. A known control method is the multiple inputs / multiple outputs (MIMO) control method, which is presented for example in the articles by Kirkeby, O., & Nelson, PA, “Digital filter design for inversion problems in sound reproduction”, Journal of the Audio Engineering Society, 47(7 / 8), 583-595, 1999, and Miyoshi, M., & Kaneda, Y., “Inverse filtering of room acoustics”, IEEE Transactions on acoustics, speech, and signal processing, 36(2), 145-152, 1988, DOI: 10.1109 / 29.1509.The "multiple inputs multiple outputs" (MIMO) acoustic field control method consists of jointly using several loudspeakers (constituting the multiple inputs) to optimally control the acoustic field at several measurement points (constituting the multiple outputs). The implementation of a MIMO method requires, in a calibration phase, measuring the crossed acoustic responses along all possible paths between each of the multiple loudspeakers and each of the multiple microphones. The MIMO method requires defining a target acoustic field ^^ consisting of each of the target acoustic responses desired for each of the measurement points.The MIMO method consists, in an optimization phase, of synthesizing digital filters from all the measurements of the cross-responses and the target acoustic field and, in a use phase, of applying these filters to the signals addressed to the loudspeakers of the sound reproduction system. The synthesis of the MIMO filters is carried out by an optimization process which consists of determining a filter for each of the loudspeakers so that the joint action of the filtered loudspeakers produces a modified acoustic field which is as close as possible to the target acoustic field ^^. In particular, the MIMO method differs from the single input – multiple output – SIMO method (“Single Input Multiple Input”), in which the control of the acoustic field is carried out by a single loudspeaker (single input of the process) and the optimization phase determines a single filter applied to this single loudspeaker.In the case of a multi-speaker system, the SIMO method is applied to each of the speakers in turn. Unlike a MIMO method, the SIMO method does not take into account the combined action of the different speakers used together since the optimization is calculated speaker by speaker separately and independently. More precisely, the MIMO control method consists of calculating optimal filters, or filters, ^^, from the target responses ^^ and a response matrix ^^. The element ℎ. ^^^^ of the matrix ^^ is representative of the impulse response (or transfer function) between the s-th loudspeaker of the sound reproduction system and the p-th microphone of a set of microphones arranged in the listening area. The elementℎ^^^^ results from a measurement carried out while the s-th loudspeaker is excited by a test signal, the other loudspeakers not emitting. We are looking for filters ^^ such that:^^ = ^^. ^^Soit : ^^ = (^^^^ . ^^)−1 ⋅ (^^^^ . ^^)However, the direct calculation of filters has several drawbacks: First of all, the matrix ^^^^ . ^^ is often not well conditioned and its inversion can cause instability problems of the calculated filters, which results in artifacts. This drawback can be solved by introducing regularization terms in the cost function ^^(^^) used for optimization. However, this causes a loss of precision in the filters, more or less strong depending on the weight of these regularization terms. If the number of loudspeakers and microphones is large, that is to say if the number of elements ℎ ^^^^in the matrix ^^ is important, the digital cost of the inversion is too high to be managed by the computer of a commercial audio equipment. High-performance inversion techniques, which take into account the structure of the matrix ^^ are then necessary. However, the number of measurement points and the number of loudspeakers constitute a limitation of the MIMO control method. If the targets ^^ do not respect certain physical constraints linked to the propagation of sound in the room, the filters ^^ obtained may present audible artifacts and / or a strong non-causal component ("pre-ringing"). This problem appears in particular when synthetic targets are imposed that are too ideal. Several approaches have recently been proposed to improve the MIMO control method, by overcoming, at least partially, all or part of these drawbacks. In particular, in patent application FR n°2300786,the applicant proposed an approach which consists of arranging the loudspeakers in two networks (“array”), the first network comprising emitting loudspeakers and the second network comprising non-emitting loudspeakers, and defining a target acoustic field which will be obtained from the emitting loudspeakers,non-emitting loudspeakers operating to cancel the reflections of the waves emitted by the emitting loudspeakers. This approach leads to a matrix ^^ and targets ^^ whose shape greatly facilitates the resolution of the problem and allows the obtaining of filters ^^ for all the loudspeakers of the sound reproduction system with very few artifacts and a very high control of the acoustic field in the listening area. The emitting loudspeakers are advantageously selected and / or placed so as to minimize the excitation of the transverse modes of the room. They are for example placed on a so-called "front" wall and possibly on a wall opposite the room, called the "rear" wall. This makes it possible to create a substantially constant wavefront over the entire listening area, minimizing the spatial variations of the acoustic field. In the calibration phase,This process also relies on time windowing to retain only the wavefront emitted by the transmitting loudspeakers. More compact digital equalization filters with fewer artifacts are then obtained, without the need for excessive regularization. The calculated filters contribute to generating a wavefront similar to the wavefront corresponding to the targets. They thus reduce reflections between the front and rear walls. However, for there to be a wavefront, a wave must propagate in the room. However, this is only possible when the wavelength is greater than approximately twice the characteristic dimension of the room. This wavelength is associated with a so-called cutoff frequency, which depends on the geometry of the room. More precisely, the cutoff frequency ^^0 corresponds to the first longitudinal mode of the room, ^^0 = 0.5^^0 / ^^^^,where ^^0 =343m / s is the speed of sound and ^^^^ is the length of the room. For example, for a room 8.5m long, the first longitudinal mode has a wavelength of 17m, or a cutoff frequency of 20Hz. Thus, the MIMO control method of document FR n°2300786 is only optimal for a sound wave propagation regime. It is not suitable for waves with a wavelength greater than the wavelength associated with the cutoff frequency. However, a sound reproduction system can emit waves in a wide frequency range, particularly below the cutoff frequency. This is particularly the case when it includes subwoofers or loudspeakers capable of reproducing low frequencies. These low-frequency sound waves contribute to the user's auditory experience. Below this cutoff frequency and in a closed room with perfectly sealed and rigid walls,the acoustic field is mainly composed of a single mode, in which the acoustic field is almost constant throughout the listening area. This is called a pressurization regime. Under ideal conditions of the pressurization regime, control of spatial variations in the acoustic field is not necessary. However, real listening rooms or rooms are not pressurization chambers with perfectly rigid and airtight walls. On the one hand, the walls of the rooms are themselves resonant membranes ("drumhead" effect) which produce a non-uniform acoustic field in the room. On the other hand, real rooms are never perfectly airtight, if only to ensure their ventilation,not to mention the many semi-open rooms. Real rooms are not pressurized chambers spontaneously producing a uniform acoustic field. It thus appears that the optimization strategy (i.e. the choice of targets to be reached) must be adapted to each frequency band of the operating frequency range of the sound reproduction system, in particular a frequency band below the cut-off frequency and a frequency band above the cut-off frequency. It would therefore be desirable to improve the method of the state of the art so that it can also allow control of the acoustic field in the listening area by frequency band, in particular not only in the frequency band above the cut-off frequency,but also in the frequency band below the cut-off frequency. The aim of the present invention is to solve this problem. For this purpose, the subject of the invention is a method for MIMO control of an acoustic field in a listening area inside a room, the acoustic field being generated by a sound reproduction system comprising a plurality of loudspeakers and a first computer, the control method being implemented by a control system connected to the plurality of loudspeakers and comprising a plurality of microphones positioned in the listening area and a second computer, the method being characterized in that it comprises the steps consisting of,in a calibration phase carried out by the second calculator: defining at least a first frequency band and a second frequency band in the operating frequency range of the sound reproduction system; defining a first filter optimization strategy on the first frequency band and a second filter optimization strategy on the second frequency band; acquiring a response matrix of the loudspeakers and of the room by applying a test signal to each loudspeaker successively and by measuring an elementary response for the loudspeaker activated at each of the microphone positioning points; and, determining a global filter as a function of the response matrix, the first strategy and the second strategy. According to particular embodiments, the control method comprises one or more of the following characteristics,taken in isolation or in all technically possible combinations: - defining a first filter optimization strategy on the first frequency band and a second filter optimization strategy on the second frequency band consists of defining a first target for the first band and a second target for the second band, and determining a global filter consists first of determining by optimization a first filter for the first band from the response matrix and the first target and a second filter for the second band from the response matrix and the second target,then combining the first and second filters to obtain the global filter. - defining a first filter optimization strategy on the first frequency band and a second filter optimization strategy on the second frequency band consists of defining a first target for the first band and a second target for the second band, and determining a global filter consists of first combining the first and second targets to obtain a global target, then determining by optimization the global filter from the response matrix and the global target. - defining a first filter optimization strategy on the first frequency band and a second filter optimization strategy on the second frequency band consists of defining a first target for the first band and a second target for the second band, and determining a global filter consists of: filtering the response matrix on the first frequency band,respectively on the second frequency band, to obtain a first response matrix, respectively a second response matrix; determining by optimization a first filter for the first band from the first response matrix and the first target and a second filter for the second band from the second response matrix and the second target; then, combining the first and second filters to obtain the global filter. 5 - defining a first optimization strategy for the filters on the first frequency band and a second optimization strategy for the filters on the second frequency band consists of defining a first target for a particular frequency of the first frequency band and a second target for a particular frequency of the second band, and determining a global target varying continuously with the frequency by, 10interpolation of the first and second targets, then determining by optimization the global filter from the response matrix and the global target. - the first target ^^ ^ ( ^ 1) ^^ at a positioning point p of a microphone is synthesized: in a first form: ^^(1)^^ = ^^0^^(^^ − ^^0), ∀^^ where A_0 is the amplitude of the acoustic field and t_0 is a reference time of emission of the test signal; in a second form: ^^(1) ^^ = ∑^^ ^^=1 ℎ^^^^^^^^^^ , ∀^^, where ℎ^^^^^^^^^^ is the response of the s-th loudspeaker measured at a single reference point, ^^ ^^^^^^ ; in a third form: ^^(^ ^ 1) = where ^^^^ is a subset of the measurement points comprising ^^^^ points, and ℎ ^^^^ is the response of the s-th loudspeaker measured at point ^^; or, in a fourth form: ^^(1) ^^ = ∑^^ ^^=1 ℎ^^^^ .20 - the second target at a point ^^ of positioning of a microphone is: in a first form: ^^(2)^^ = ^^^^^^(^^ − ^^^^), where ^^ is the Dirac function, ^^^^ is the amplitude of the wave at the measurement point ^^, and ^^ ^^ is the delay which depends on the measurement position relative to the or each loudspeaker emitting the test signal; under one second f orme : ^^(^^) is a time window applied to the 25 ℎ^^^^(^^) measured, and ^^^^ is the subset of loudspeakers involved in the creation of the target; in a third form: ; or, in a f orme : , where ^^ is the distance between the loudspeaker of index ^^ and the measurement point ^^. - the part defining a cut-off frequency between a pressurization regime and a propagation regime, the first frequency band is defined below the cut-off frequency, and the second band is defined above the cut-off frequency. The invention also relates to a control system adapted to implement, when associated with a sound reproduction system, the preceding MIMO control method. Preferably, the control system comprises a first computer and a plurality of microphones and the sound reproduction system comprises a second computer and a plurality of loudspeakers, the first and second computers being the same suitably programmed computer or two separate computers but connected for implementing the MIMO control method. The invention also relates to a computer program comprisingsoftware instructions which, when executed by a computer, implement the preceding control method. The invention and its advantages will be better understood upon reading the following detailed description of a particular embodiment, given solely as a non-limiting example, this description being made with reference to the appended drawings in which: [Fig 1] Figure 1 is a schematic representation of an embodiment of a system for controlling an acoustic field according to the invention; [Fig 2] Figure 2 is a representation in the form of functional modules of a first programming mode of the computer of Figure 1 (and of a first embodiment of the control method according to the invention); [Fig 3] Figure 3 is a representation in the form of functional modules of a second programming mode of the computer of Figure 1 (and of a second embodiment of the control method according to the invention); [Fig4] Figure 4 is a representation in the form of functional modules of a third mode of programming the computer of Figure 1 (and of a third embodiment of the control method according to the invention); [Fig 5] Figure 5 is a representation in the form of functional modules of a fourth mode of programming the computer of Figure 1 (and of a fourth embodiment of the control method according to the invention); [Fig 6] Figure 6 represents several graphs, each graph giving the amplitude of the processed response as a function of the frequency for different measurement points of the listening area, in order to allow a comparison between the implementation of the method according to the invention and that of the method according to the state of the art; and, [Fig 7] Figure 7 represents different diagrams for defining the subset of the transmitting loudspeakers for as many frequency bands subdividing the operating frequency range of the system ofsound reproduction. GENERAL INFORMATION The method for MIMO control of an acoustic field according to the invention provides for subdividing the operating frequency range of the sound reproduction system into at least two frequency bands, and defining an optimization strategy on each of the bands individually, so as to be able to take into account filter optimization parameters which vary with the frequency. In particular, the frequency range is subdivided into a first low-frequency band, below the characteristic cut-off frequency of the room, and a second complementary band, or medium / high-frequency band, above the cut-off frequency. In this way, an optimization strategy dedicated to the pressurization regime leads to the optimization of the filters specifically on the first band and an optimization strategy dedicated to the propagation regime leads to the optimization of the filters specifically on the second band.filters are thus introduced to re-establish a uniform pressure field for the pressurization regime. It is relevant to implement a MIMO process because it exploits the combined action of the different loudspeakers through specially optimized filters to optimally re-establish a uniform acoustic field in a non-ideal real room. Moreover, even under ideal conditions, the use of a MIMO process remains relevant because the loudspeakers do not necessarily all have the same characteristics or the same capacities to produce the acoustic pressure in the listening area. Thus, even if the target acoustic field d is uniform, the optimal filters g are never simple identity filters (or "bypass") because the MIMO process adapts to the precise characteristics of each loudspeaker and more generally to the acoustic details of the measured responses contained in matrix H. Finally, although in the ideal case of the pressurization regime theacoustic field does not present any natural spatial variation, the MIMO method according to the invention makes it possible to introduce controlled variations in order to attenuate or accentuate this frequency band in certain areas of the room. The choice of a non-uniform target acoustic field as well as the use of a MIMO method jointly using the different loudspeakers and the different measurement points makes it possible to obtain an acoustic field presenting controlled spatial variations. This technique makes it possible, for example, to introduce personalized equalizations according to the preferences of the different listeners present in the room. Advantageously, the complementary band is itself subdivided into several frequency bands, for example a medium frequency band and a high frequency band, an optimization strategy specific to each band then being implemented. The optimization parameters which vary with frequency are, for example: -physical parameters, which have an influence on the propagation of sound, such as the geometry and acoustic absorption of walls. These parameters determine the shape and degree of complexity of the acoustic field that one wishes to optimize; - subjective parameters, linked to the perception of sound, which may make it desirable to accentuate certain frequency intervals despite a flat response; - hardware-related parameters, such as the number of loudspeakers and microphones; - digital parameters, such as the number of interpolation points of each filter, the duration of each filter, the compactness of each filter, the dynamics of each filter. FIGURE 1 Figure 1 generally represents a system for controlling an acoustic field generated by a sound reproduction system. The control system 10 is configured to control the acoustic field in a listening area 19, preferably three-dimensional. The listening area 19 isentirely located inside a room 20, such as a movie theater, a concert hall, or the equivalent. Alternatively, the listening area 19 is identical to the volume of the room 20. In the present embodiment, the control system 10 is associated with a sound reproduction system 11. The sound reproduction system 11 comprises a set of ^^ loudspeakers 12 and a computer 18. The computer 18 is for example integrated into an audio amplifier. Each loudspeaker 12 is connected to a dedicated output of the computer 18 by a dedicated link, wired or non-wired. In the present embodiment, the set of loudspeakers comprises a plurality of loudspeakers 12 each operating over the entire frequency range. But, generally speaking, it is sufficient to have at least one loudspeaker operating at least in a first frequency band and at least one loudspeaker operating at least in a second frequency band. For example, the geometryof the listening room 20 makes it possible to know the cut-off frequency. This cut-off frequency is then advantageously taken into account to define a transition frequency around the cut-off frequency which delimits the first frequency band and the second frequency band in the frequency range of the sound reproduction system 11. The first frequency band corresponding to a pressurization regime and the second frequency band corresponding to a propagation regime. A sound source 9 is connected to the input of the computer 18. This is for example a compact disc player, a record player, a multimedia server, a preamplifier, or the equivalent. The computer 18 is suitably programmed to, during the use phase of the sound reproduction system 11, apply a set of filters ^^ to the raw electrical signal^^0, received from the sound source 9. The filter set ^^ comprises a filter per loudspeaker of the sound reproduction system. The computer 18 outputs a set of filtered signals ^^ ^^, each filtered signal being transmitted to the associated loudspeaker 12 so that it generates a sound wave. The superposition of the sound waves generated by the different loudspeakers in the listening area 19 defines the acoustic field. The control system 10 is adapted to control the sound reproduction system 11. It comprises a plurality of ^^ sensors 16 and a control computer, which in the present embodiment is advantageously the computer 18. The sensors 16 are microphones, preferably identical to each other. The sensors 16 are preferably omnidirectional microphones, the sensitivity of which does not depend, or very little, on the direction of incidence of the acoustic wave on the microphone. The sensors 16 are prepositioned at different points ^^ of the listening area 19. Each sensor defines a measurement position of the sound waves generated by the loudspeakers 12.Each sensor 16 is connected to an input of the controller 18 by a dedicated link, wired or non-wired. The computer 18 is suitably programmed to, in the calibration phase of the sound reproduction system 11, calculate the filters ^^ which will be applied in the use phase of the system 11. More precisely, the computer 18 comprises calculation means, such as a processor 32, and storage means, such as a memory 30, as well as an input / output interface 34, for the connection of the source 9, the loudspeakers 12 and the sensors 16. The memory of the controller 18 stores in particular the instructions of computer programs, in particular a program 38, the execution of which in the use phase makes it possible to apply filters ^^ to the raw signal ^^0 before applying a filtered signal ^^. ^^to each loudspeaker 12, and a program 36, the execution of which in the calibration phase allows the implementation of the MIMO control method according to the invention to optimally determine the filters ^^. The execution of these programs provides the controller 18 with several functionalities, represented in the form of functional modules in the figures. 1 er EMBODIMENT In a first embodiment, illustrated in FIG. 2, the controller 18 is programmed to comprise a control module 102 and a filtering module 104. The control module 102 comprises: - a unit 112, for delimiting a plurality of ^^ frequency bands within the operating frequency range of the sound reproduction system; - a unit 114 for acquiring the response matrix ^^; - a unit 116 for determining the targets for each frequency band ^^ of the plurality of frequency bands; -a plurality of units 118b for calculating the filters ^^(^^) for each frequency band^^ of the plurality of frequency bands, from the response matrix ^^ and the targets ; and, - a 120 filter aggregation unit ^^ (^^) in global filters ^^. The filtering module 104 of the computer 18 is configured with the global filters ^^ at the end of the calibration phase and, in the use phase, it is adapted to apply these global filters ^^ to the raw signal ^^0 received from the source 9, in order to obtain a set of filtered signals ^^ to drive the loudspeakers, the ^^-th filtered signal ^^ ^^of this set being applied to the ^^-th loudspeaker 12 of the system 11. The method implemented by the computer is then as follows: In a calibration phase 101, a number ^^ of sensors 16 and a positioning of each of these sensors in the listening area are defined. Thus, ^^ loudspeakers and ^^ sensors are used for the calibration. Then, the module 102 is executed so that: - In a step 111, the unit 112 is executed to delimit frequency bands. The operating frequency range of the sound reproduction system 11 is thus subdivided into ^^ frequency bands. Each band is indexed by an integer ^^ between 1 and ^^. This delimitation is carried out for example by selecting the transition frequency which defines the boundary between two consecutive bands, ^^ and ^^+1. Alternatively, the consecutive bands partially overlap over an interval around a transition frequency.In a simple embodiment, the system 10 comprises a human-machine interface allowing an operator to enter the number ^^ of bands and the values ​​of the transition frequencies ^^. (^^) . In particular, the operator can measure, by dedicated means, the cut-off frequency of the room in which the sound reproduction system 11 is installed. He can also know the characteristics of the loudspeakers 12 and their respective frequency operating domains. Such information allows him to define the frequency band configuration of the system 11. - In a step 113, the unit 114 is executed to obtain the response matrix ^^ of the sound reproduction system 11 and of the room 20. The matrix ^^ is for example a multichannel convolution matrix for the ^^ loudspeakers 12 and the ^^ sensors 16. It is for example defined per block as follows: The element ℎ ^^^^is an elementary convolution matrix associated with the measurement carried out by the p-th sensor 16 of the wave emitted by the s-th loudspeaker 12, while the latter is excited by a test signal ^^ ^^^^^^^^ (the other loudspeakers remaining silent). More precisely, the element ℎ^^^^ is associated with the measurement during a period ^^ℎ. ∆^^ of the sound wave emitted by the emitting loudspeaker (∆^^ being the sampling time step and ^^ ℎ a predefined integer). The element ℎ ^^^^ is for example a Toeplitz matrix, of the form: where ℎ ^^^^ (^^) is the amplitude measured at sampling time t between 1 and ^^ ℎ . - In a step 115, the unit 116 is executed to define targets for each band ^^. The targets for the band ^^ correspond to the control strategy chosen for this band. Preferably, the control strategy is different from one band to another. Advantageously, as explained in application FR No. 2300786, a subset of all the loudspeakers of the system 11 is chosen as transmitting loudspeakers to create the targets. To define the targets different from one band to another, the subset of transmitting loudspeakers changes according to the band considered. The targets take for example the form of a column vector: Each component ^^ ( ^ ^ ^^) , or target at point ^^, represents the target response at point ^^, with ^^ an integer between 1 and ^^, ^^ being the number of measurement points where the microphones are placed 16. Each component ^^ ( ^ ^ ^^) is a byte of size ^^ ^^representing the impulse response of the target acoustic field, in the frequency band ^^, at point ^^ and for each sampling instant t between In the following, an example of a strategy adapted to a pressurization regime is first presented, and an example of a strategy adapted to a propagation regime is then presented. In the pressurization regime, the targets are determined. The validity domain of these targets is between the zero frequency and a first transition frequency ^^ (1) around the frequency of the first longitudinal mode of room 20, from which the acoustic field is no longer constant at all points in the room. The targets are synthesized to correspond to a uniform acoustic field constant throughout the room: where ^^0 is the amplitude of the desired or sought acoustic field at the positioning point p of the measuring sensor and ^^0 is the reference time of emission of the test signal ^^ ^^^^^^^^ . It may be noted here that although the targets are constant, in the case of a non-ideal room and loudspeakers, the filters obtained by the MIMO method will not be identity filters. These targets may advantageously be band-pass or low-pass filtered to ensure that their spectra coincide with the bandwidth of the loudspeakers 12. Alternatively, the targets of the pressurization regime are synthesized to correspond to a non-uniform acoustic field throughout the room: where ^^ ^^ is the amplitude of the desired or sought acoustic field at the positioning point p of the measurement sensor and ^^0 is the reference time of emission of the test signal ^^ ^^^^^^^^. By "amplitude of the acoustic field" we mean the amplitude of the acoustic wave at this point and not a simple measurement of power at this point. Alternatively, the targets of the pressurization regime are determined from the measurement of the system responses. Thus, in a first possible implementation, the target at point ^^ is defined as the sum of the sound waves emitted by each loudspeaker and measured at a single point, called the reference point, ^^ ^^^^^^ , of listening area 19: This first implementation remains valid in situations where the real acoustic field in the listening area is very homogeneous below the transition frequency ^^ (1). In this case, the targets are weakly dependent on the measurement point and measurements at a single point in the listening area are sufficient to represent the sound field in the entire listening area. In a second possible implementation, the target at point ^^ is also unique for all measurement points, but this time it is obtained as the average response measured at at least two different measurement points: where ^^ ^^ is a subset of the measurement points comprising ^^ ^^points. This spatial average allows to define targets more robust to variations in the acoustic field, variations which can occur in real situations, due for example to the presence of large obstacles in the room, or by the use of loudspeakers with strong differences in sensitivity and / or frequency response. In a third possible implementation, the target at point ^^ is different for each measurement point ^^: These targets allow for greater variations in the measured acoustic field to be taken into account compared to the theoretical constant acoustic field. Regardless of the chosen implementation, additional processing can be applied to the targets ^^ (1) , such as time windowing, or time or frequency smoothing, in order to reduce the filtering effort and thus improve the efficiency of the inversion algorithm. In the propagation regime, the targets ^^ (2)are, in a first implementation, constructed so as to synthesize a plane wave propagating from a wall of the room chosen as the front wall (and the loudspeakers arranged on this wall as emitters, the other loudspeakers of the system 11 being non-emitters). The amplitude of the plane wave being constant over the transverse section of the room, for each measurement point ^^, the target response at point ^^, ^^(2)^^, is defined as: where ^^ is the Dirac function, ^^ ^^ is the amplitude of the wave at the measurement point ^^ and ^^ ^^ is a delay, which depends on the measurement point relative to the transmitting loudspeaker. This delay can be defined as: ^^^^ = ^^0 + ^^^^ / ^^0, where ^^0 is the speed of sound in the air of the room and ^^ ^^is the distance between the front wall and the measurement point. The reference time ^^0 must be chosen so that the target field respects the causality condition, (^^ − ^^^^) ≥ 0, for all loudspeakers used as emitter in the calibration phase. The amplitude ^^ ^^ can be chosen to be constant for all points ^^ (case of a room without losses), or variable according to the position of the point ^^ (case of a room with losses). As before, the targets can be band-pass or low-pass filtered so that their spectra coincide with the bandwidth of the emitting loudspeakers. Since the targets are composed only of the transverse plane mode of the room, any contribution from higher-order modes is eliminated. This has the effect of minimizing the spatial variations of the acoustic field. The range of validity of these targets is between the cut-off frequency and a second transition frequency ^^ (2). The latter is equal to the spatial aliasing frequency, determined by the spacing of the measurement points. Typically, ^^ (2) is chosen by arranging the microphones 16 so that the associated wavelength is at least twice as large as the spacing between the measurement points, ^^^^. This avoids ambiguity in the measurement of the acoustic field by spatial aliasing. By taking a spacing ^^^^ equal to one third of the wavelength associated with ^^ (2), the second crossover frequency is then given by ^^(2) = ^^0 / (3^^^^). This frequency must be high enough to control as many modes as possible, but small enough not to require too many measurement points to correctly sample the sound field. A good compromise is to take ^^^^ ≈ 1 m, which gives ^^(2) ≈ 100 Hz. The range of validity of the targets is also linked to the spacing between the loudspeakers. For a two-dimensional loudspeaker array, the aliasing frequency is determined by ^^(2) = with ^^^^ and ^^^^ the number of loudspeakers in directions ^^ and ^^ respectively and with ^^^^ and ^^^^ the room dimensions in directions ^^ and ^^ respectively. In a second implementation, the target at point ^^, ^^ ( ^ ^ 2) , of the propagation regime is determined from the measurement of the responses of the real system: where ^^ ^^is the subset of the emitting loudspeakers and ^^(^^) is a time window that allows to keep only the contribution of the first wavefront. These targets guarantee a good homogeneity of the acoustic field while implicitly respecting the physical constraints of the propagation of the sound in the room, such as the arrival time and the amplitude decay of the wave during its propagation. - In a step 117, the different units 118 (b) are executed independently of each other, advantageously in parallel, in order to determine the filters ^^ (^^) for each band ^^. A vector of filters contains ^^ optimization filters ^^ ^^ ∶ Each component ^^ ^ ( ^ ^^) , or speaker filter s, is a byte of duration ^^ ^^ .∆^^ representing the impulse response of the filter associated with the loudspeaker ^^ for each sampling instant t between 1 and ^^^^ : The MIMO control method according to the invention consists of finding, for each band^^ considered independently, the filters which verify the equation:^^. ^^(^^) = ^^(^^)The vector of filters is the result of an optimization. For example, filters are calculated in the least squares sense by minimizing the cost function ^^ :^^ (^^(^^)) = ||^^ ∙ ^^(^^) − ^^(^^) ||2 + ^^ . ^ (^^) (^^)1 ^^^^^1 (^^ , ^^) + ⋯+ ^^^^ . ^^^^^^^^ (^^ , ^^)where the functions ^^^^^^ ^^ , r integer between 1 and ^^, ^^ being the number of regularization functions, are regularization functions, which depend on one or more regularization parameters ^^, is a regularization coefficient. For the case without regularization function (^^^^ = 0, ∀^^), the filters sont par example defined as follows: For the general case, we can write the following relation: where ^^ is the inversion function that minimizes the cost function ^^ when it contains one or more regularization terms. - Finally, in a step 119, the unit 120 is executed to determine the global optimization filters ^^ by combining the filters ^^ (^^) optimized on each band. In a simple embodiment, this combination is done by means of a frequency crossover on an interval around each transition frequency ^^ (^^)between adjacent optimization bands. The type of overlap can be freely chosen depending on the response that is desired on either side of the transition frequency. The slope of the filters in the overlap interval is adapted to obtain a more or less smooth transition between the adjacent bands. In a phase 103 of use which follows the calibration phase 101, the unit 104 is executed to filter the signal ^^0 by applying the global optimization filters ^^ to it. A filtered signal ^^ ^^ is obtained for each loudspeaker ^^ of the system 11. 2 dEMBODIMENT In a second embodiment, illustrated in FIG. 3, the controller 18 is programmed to comprise a control module 202 and a filtering module 104. The control module 202 comprises: - a unit 112, for delimiting frequency bands within the operating frequency range of the sound reproduction system 11; - a unit 114 for acquiring the response matrix ^^; - a unit 116 for determining the targets for each frequency band^ ^ ; - a 220 target aggregation unit to define global targets ^^. - a unit 218 for calculating the global filters ^^ from the matrix ^^ and the global targets ^^. The filtering module 104 of the calculator 18 is configured with the global filters ^^ at the end of the calibration phase and, in the use phase, it is adapted to apply these global filters to the raw signal ^^0 received from the source 9, in order to obtain a plurality of filtered signals ^^ ^^to drive each of the loudspeakers 12. The method implemented by the computer is then as follows: In a calibration phase 201, a number and an arrangement of the set of loudspeakers 12 and a number and a positioning of the set of sensors 16 is determined. Then, the module 202 is executed so that: - In a step 211, identical to step 111, the unit 112 is executed to delimit frequency bands. - In a step 213, identical to step 113, the unit 114 is executed to obtain the response matrix ^^ of the sound reproduction system 11 and the room 20. - In a step 215, identical to step 115, the unit 116 is executed to define targets ^^ (^^) for each band ^^ and advantageously the subset of transmitting loudspeakers to create this target. - In a step 216, the unit 220 is executed to aggregate the targets ^^ (^^)for each of the bands b, in global targets ^^ for the entire operating range of the sound reproduction system 11. - In a step 217, the unit 218 is executed in order to determine the global filters ^^ from the global targets ^^ and the matrix ^^. In step 217, the global filters are optimized, the solution of the equation: ^^. ^^ = ^^In general, the following relationship can be written: ^^ = ^^(^^, ^^, ^^)In a phase 203 of use which follows the calibration phase 201, the unit 104 is configured with the global filters ^^ and executed to filter the signal ^^0 by applying the global optimization filter ^^ to it. A filtered signal ^^ ^^ is obtained for each loudspeaker 12 of the system 11. 3 eEMBODIMENT In a third embodiment, illustrated in FIG. 4, the controller 18 is programmed to comprise a control module 302 and a filtering module 104. The control module 302 comprises: - a unit 112, for delimiting frequency bands within the frequency range of the sound reproduction system; - a unit 114 for acquiring the response matrix ^^; - a unit 116 for determining targets for each frequency band ^^; - a unit 322 for determining a response matrix by band ^^; - a plurality of units 318 b of filter calculation for each band b, from the matrix ^^ (^^) and targets and, - a 120 filter aggregation unit ^^ (^^)in global filters ^^. The filtering module 104 of the computer 18 is configured with the filters ^^ at the end of the calibration phase and, in the use phase, it is adapted to apply this global filter to the raw signal ^^0 received from the source 9, in order to obtain filtered signals ^^ ^^to drive each loudspeaker. The method implemented by the computer is then as follows: In a calibration phase 301, a number and an arrangement of the set of loudspeakers 12 and a number and a positioning of the set of sensors 16 is determined. Then, the module 302 is executed so that: - In a step 311, identical to step 111, the unit 112 is executed to delimit frequency bands. - In a step 313, identical to step 113, the unit 114 is executed to obtain the response matrix ^^ of the sound reproduction system 11 and of the room 20. - In a step 315, identical to step 115, the unit 116 is executed to define targets for each band ^^ and, advantageously, the subset of transmitting loudspeakers for each band ^^. - In a step 316, the unit 322 is executed to determine matrices per band ^^ from the list of crossover frequencies ^^ and the matrix ^^, for example by applying a bank of low-pass, band-pass or high-pass filters, whose cut-off frequencies are defined from the transition frequencies . - In a step 317, each unit 218 b is executed to determine the filters ^^ (^^) associated with the band ^^ from the targets and the matrix for the band ^^. In step 317, the filters ^^(^^) are optimized, solution of the equation:^^(^^). ^^(^^) = ^^(^^)Generally, the following relation can be written:^^(^^) = ^^(^^(^^), ^^(^^), ^^)-Finally, in a step 319, the unit 120 is executed to determine the global filters^^ by combining the filters ^^(^^) optimized on each of the bands ^^.In a phase 303 of use which follows the calibration phase 301, the unit 104 is executed to filter the signal ^^0 by applying the global filters ^^ to it. A filtered signal ^^ ^^is obtained for each loudspeaker ^^ of the system 11. 4 e EMBODIMENT In a fourth embodiment, the determination of the targets ^^ is carried out by varying the optimization parameters continuously as a function of the frequency instead of using optimization parameters specific to each band and therefore constant on each frequency band. This fourth embodiment is illustrated in FIG. 5. The controller 18 is programmed to comprise a control module 402 and a filtering module 104. The control module 402 comprises: - a unit 112, for delimiting frequency bands within the frequency range of the sound reproduction system; - a unit 114 for acquiring the response matrix ^^; - a unit 416 for determining the global targets ^^, comprising a block 452 for determining a response matrix of the emitters H E, a block 454 for calculating a transformed response matrix of the emitters H ET and a block 456 of target synthesis at each measurement point ^^ ^^ ; and, - a unit 218 for calculating the global filters ^^ from the matrix ^^ and the global targets ^^. The filtering module 104 of the calculator 18 is configured with the filters ^^ at the end of the calibration phase and, in the use phase, it is adapted to apply these filters to the raw signal ^^0 received from the source 9, in order to obtain filtered signals ^^ ^^to drive each loudspeaker. The method implemented by the computer is then as follows: In a calibration phase 401, a number and an arrangement of the set of loudspeakers 12 and a number and a positioning of the set of sensors 16 is determined. Then, the module 402 is executed so that: - In a step 411, identical to step 111, the unit 112 is executed to delimit frequency bands. - In a step 413, identical to step 113, the unit 114 is executed to obtain the response matrix ^^ of the sound reproduction system 11 and of the room 20. - In a step 415, the unit 416 is executed to define global targets ^^. More precisely, in a sub-step 451, the block 452 is executed to define transmitting loudspeakers for each frequency band. A filter ^^ ^^ (^^) is then determined which represents the contribution of the speaker with index ^^ to the creation of the targets as a function of frequency. The filter ^^^^ ( ^^ ) applies to each of the elements ℎ ^^^^ of the H matrix. For example, in the case of the second frequency band corresponding to the propagation regime, only the loudspeakers on the front wall are emitting. In the middle of this second band, the filter ^^ ^^ (^^) is 1 for transmitting loudspeakers and 0 for non-transmitting loudspeakers. In the case of the first frequency band corresponding to the pressurization regime, all loudspeakers are transmitting. In the middle of this first band, the filter C s (f) is 1 for all speakers. The intermediate values ​​of the filter ^^ ^^ ( ^^ ) result from an interpolation according to the frequency of these fixed values. The value of the filter ^^ ^^ ( ^^ )for non-transmitting loudspeakers the propagation mode gradually changes from 0 to 1 as the frequency decreases to move from one regime to the other. The matrix ^^ is then filtered with the filter ^^ ^^ (^^) to obtain a filtered matrix ^^ ^^ . It is called the transmitter response matrix. For each measurement point ^^ and each loudspeaker ^^, we have:ℎ ^^ ^^ ^^ = ℎ^^^^ . ^^^^(^^)In a sub-step 453, block 454 is executed to transform the responses of the transmitting loudspeakers by applying a time window.In a first step, a time window specific to each frequency band^^ is defined. Then, in a second step, an interpolation according to the frequency of these time windows is carried out to define an identical time-frequency transformation ^^ to be applied to each of the elements ℎ ^ ^ ^ ^ ^^ of the matrix ^^ ^^and thus obtain a transformed response matrix ^^ ^^^^ for the transmitting loudspeakers. For each point ^^ of measurement and each loudspeaker ^^, we have:ℎ ^^^^ = ^^^^^^ ^^(ℎ^^^^ )This transformation allows to select the part of the measured response that is used to develop the target. A narrow windowing interval allows to take into account only the first wavefront emitted by the loudspeakers, while a wider windowing interval allows to take into account also the reflections on the walls of the room. For example, in the case of the second optimization band specific to the propagation regime, only the first wavefront is selected excluding the room. In the middle of this second band, the ^^ transformation corresponds to a fairly short time window, typically shorter than the propagation time to cross the room. In the case of the first optimization band specific to the pressurization regime, the whole response must be taken into account to maximize the pressurization. In the middle of this first band, the ^^ transformation corresponds to a very wide time window.The lengths of the windowing interval are interpolated as a function of the frequency between these fixed values. This amounts to performing a windowing whose width depends on the frequency. In a substep 455, block 456 is executed to determine the targets ^^. In the simplest embodiment, the target at point ^^ is the combination of the transforms of the filtered responses at point ^^ : ^. ^ ^^^^ = ∑ℎ^^ ^^ ^^ ^^ ^^=1 In more advanced embodiments, the target at the measurement point ^^ is achieved by applying to each loudspeaker ^^ a time-frequency weighting ^^^^: This more advanced mode makes it possible to compensate for position errors of the transmitting loudspeakers and / or to optimize the wavefront synthesis in the propagation regime. Finally, in a step 417, the unit 418 is executed in order to determine the filters^^ as:^^ = ^^(^^, ^^, ^^) In a phase 403 of use which follows the calibration phase 401, the unit 104 is executed to filter the signal ^^0 by applying to it the global optimization filter ^^. A filtered signal ^^ ^^is obtained for each loudspeaker ^^ of the system 11. With this fourth embodiment, a time / frequency window is used to define an optimization strategy that is continuously variable with the frequency. This constitutes an advantage in terms of flexibility compared to the other embodiments. EXAMPLE 1 An example of application of the method according to the invention is presented in relation to Figure 6. Consider a parallelepiped room with dimensions ^^^^ × ^^^^ × ^^^^ = 7.2^^ × 4.7^^ ×3^^, equipped with ^^ = 8 loudspeakers 12, including 4 emitting loudspeakers located on a front wall and the 4 non-emitting loudspeakers located on the rear wall. The loudspeakers are placed regularly on the front and rear walls, so as not to excite only the planar mode in the room up to a certain frequency.We define two transition frequencies that delimit two frequency bands: - In the first band, the optimization strategy consists in maximizing the response in the lowest part of the spectrum in order to obtain a richer and more powerful response (pressurization mode) and it is applied between the zero frequency and a frequency close to the cutoff frequency, typically ^^(1) = 0.8. ^^^^ or ^^(1) = 0.9. ^^^^. The frequency ^^^^ can be estimated beforehand according to the length of the room as ^^^^ = 0.5^^0 / ^^^^. Thus, ^^(1) is the frequency below which the sound field is constant at all points in the listening area. - In the second band: the sound field is strongly influenced by the resonance modes of the room. The optimization strategy then consists in minimizing the spatial and frequency deviation of the responses created by the excitation of these modes.The target field is then defined as the sum of the responses of the transmitting loudspeakers, time-windowed to take into account only the first wavefront. This mode of operation is applied between the first transition frequency and a second transition frequency related to the spacing between the microphones or loudspeakers. Typically ^^(2) = 100 ^^^^. The acoustic field is measured at a set of ^^ = 25 points separated by a maximum distance of ^^^^ = 1^^. The responses between all points and all loudspeakers are measured and assembled in the matrix ^^. The targets are then defined for each band: - For the first band: as the objective is to maximize the response in this band, the target responses are defined as the sum of the responses of all the loudspeakers. - For the second band: the target responses are obtained as the sum of the initial responses of the transmitters only.As the objective is to eliminate the contribution of higher modes, a time window is applied to the initial responses, in order to keep only the contribution of the first wavefront emitted by the emitters. For each band, the optimization filters are calculated from the target responses. The filters are combined to obtain global filters. The filters are applied to each loudspeaker to obtain the filtered frequency responses ^^. ^^ at each point ^^, which are calculated as: where ^^ ^^ (^^) and ^^ ^^^^ (^^) are the Fourier transforms of the filter ^^ ^^ and the answer ℎ ^^^^ Figure 6 shows the modulus in decibels (in dB) of the responses ^^ ^^ as a function of frequency (in Hz) for all measurement points. Graph A shows the processed responses when applying the pressurization regime strategy over the entire frequency range. We observe that, below the frequency ^^(1) , we obtain the desired behavior, namely an amplification of the responses while maintaining good spatial homogeneity. However, for frequencies higher than ^^ (1) , we observe strong spatial variations as well as a strong peak in the response around 53 Hz which is due to the excitation of one of the resonance modes of the room. The graph ^^ shows the processed responses when we apply the propagation regime optimization strategy over the entire frequency range. This strategy allows to control the behavior of the room between ^^ (1) and ^^ (2) , in particular by completely eliminating the resonance peak at 53Hz from graph A and obtaining flat and relatively homogeneous responses for all measurement points. However, no amplification is observed at frequencies lower than ^^ (1). Graph C shows the responses processed by merging the two previous optimization strategies using the present MIMO control method. At frequencies lower than ^^ (1) we observe a strong amplification of the responses, similar to the behavior obtained on graph A, while for frequencies between ^^ (1) and^^ (2) , excellent modal control is obtained. Remarkably, the spatial variations with the method according to the invention are smaller than those obtained using a single optimization strategy over the entire frequency range, namely the propagation regime strategy. This behavior is attributed to the fact that the propagation regime is applied only in the second band, between the frequencies ^^ (1) and ^^ (2), which reduces the filtering effort and improves the performance of the inversion algorithm in this band. EXAMPLE 2 A second example subdividing the frequency range into four frequency bands is presented in connection with Figure 7. Consider a parallelepiped room 20, the front wall (yz plane) of which is provided with a two-dimensional arrangement of 14 loudspeakers 12. The purpose of this example is to show how the proposed approach can be used to extend acoustic control to higher frequencies by a more judicious and efficient use of the loudspeakers 12 of the system 11. The desired target response in each of the four bands subdividing the frequency range is illustrated in each of the diagrams A to D respectively. In each diagram, the black circles represent the set of loudspeakers selected as emitters for the band considered and which consequently contribute to creating the target in the band ^^ considered, while the white circles represent non-emitting loudspeakers, which are not used to create the target response in the band ^^. Figure 7 shows only the loudspeakers that can have a different role between the different optimization bands. However, the present implementation example can also include other loudspeakers with a purely non-emitting role. These loudspeakers can be placed for example on the back wall, as described in the previous example. These non-emitting loudspeakers are not shown in Figure 7 for the sake of simplicity. The optimization strategy for the first two bands is similar to that of the previous example. In the first band (scheme A) we want to obtain a homogeneous and high amplitude response up to a frequency ^^ (1)close to the frequency of the first mode of the room. For this, the first band uses all the 12 transmitting loudspeakers simultaneously. It is also possible to use a smaller subset of loudspeakers depending on the amplitude of the desired target. In the second band, a second subset of 12 transmitting loudspeakers is used. The transmitting loudspeakers are also spaced apart in order to generate a propagating plane wave (diagram B). Unlike the previous case, we now consider a sufficiently fine mesh of measurement points, so that the maximum frequency in this second band is now given by the spacing in ^^ and ^^ between the transmitting loudspeakers. Indeed, the positioning of the loudspeakers makes it possible to create a plane wave up to a certain frequency ^^ (2), from which the loudspeakers also excite higher order modes. The creation of a flat wavefront is then no longer possible. This frequency is given by: ^^(2) = with a number of loudspeakers in the directions ^^, ^^ ^^equal to four and a number of loudspeakers in the directions ^^,^^^^ equal to two, and ^^^^ and ^^^^ the dimensions of the front wall in the directions ^^ and ^^respectively. If one wanted to achieve good acoustic control using this same optimization strategy, it would be necessary to increase the density of the loudspeaker array in ^^ and ^^, which would result in an excessively high number of loudspeakers. To overcome this limitation, the optimization strategy in the third band consists of using only columns of transmitting loudspeakers. Advantageously, these columns of transmitting loudspeakers are located near the edges of the front wall, which makes it possible to use the room as a waveguide in order to create a cylindrical wavefront without interference and whose amplitude is constant throughout the listening area (diagram C). The frequency limit of this strategy is given by: ^^(3) = lenumber of loudspeakers in a column (oriented along the z direction). Thus, it is possible to obtain acoustic control of the room at higher frequencies without needing to use too many loudspeakers. To obtain acoustic control at even higher frequencies, the optimization strategy in the fourth frequency band consists of using a single transmitting loudspeaker. Advantageously, the single transmitting loudspeaker is located in a corner of the room (diagram D). The positioning of the transmitting loudspeakers makes it possible to use the room again as a waveguide in order to create, this time, spherical waves, also free from interference by reflection on the walls. The combination of the different optimization strategies to obtain the optimization filters ^^, as well as the precise definition and processing of the targets can be carried out according to one or other of the embodiments presented above.For example, target responses for cylindrical or spherical waves can be obtained from ℎ measurements. ^^^^ thanks to the formula proposed previously: with ^^ = 3 for cylindrical waves and ^^ = 4 for spherical waves, le sous- set of loudspeakers used to create these waves, and ^^(^^) a time window that allows only the contribution of the direct field to be kept before the arrival of reflections caused by the opposite walls. Targets can also be defined synthetically from analytical solutions of the two-dimensional or three-dimensional acoustic wave equation. For example, for 2D cylindrical waves: So for example again for spherical waves in 3D: In these relations, ^^0 is a reference amplitude and ^^ ^^^^is the distance between the transmitting loudspeaker ^^ and point p. VARIANTS In an alternative embodiment, the transition frequencies are not defined by the user, but determined automatically from the dimensions of the part ^^^^, ^^^^ and ^^^^, in particular by using the formulas presented above. The automatic determination of the responses can also be based on the calculation of the performance of each optimization strategy as a function of the frequency, in order to empirically know the domain of validity of each strategy in a concrete case. It can also be based on an iteration of the calculations, the performance of each optimization strategy obtained at the end of an iteration making it possible to adapt the value of the transition frequencies for the next iteration and this until a performance criterion is verified. The initialization of the values ​​of the transition frequencies at the beginning of this iterative process can for example be done from the dimensions of the room and the number of loudspeakers used. If in the previous embodiments, the computer 18 is programmed to carry out both the calibration phase and the use phase, alternatively two different computers are used, each computer being dedicated to carrying out a particular phase. At the end of the calibration phase, the filters calculated by a calibration computer of the control system are downloaded into the resident computer of the sound reproduction system for the use phase. It should be noted that the control system must be connected to the sound reproduction system for the calibration phase in order to be able to apply the test signal to the loudspeakers of the sound reproduction system.The control system is then, for example, a set of microphones that are deployed on the site to be controlled and a laptop that is connected to the sound reproduction system. If in the previous embodiments, a software implementation of the invention has been presented, alternatively, all or part of the modules, units and / or blocks can be implemented in hardware form, for example in the form of an FPGA. ADVANTAGES Although MIMO control methods have demonstrated their effectiveness in controlling an acoustic field, the present invention makes it possible to define different optimization strategies depending on the frequency. Compared to the state of the art which only provides a single control strategy, the invention makes it possible to relax the constraints in the calculation of the optimization filters and to avoid certain audible artifacts.In addition, the use of several optimization strategies allows to fully and efficiently exploit all the loudspeakers of the sound reproduction system.

Claims

CLAIMS 1. Method for MIMO control of an acoustic field in a listening area (19) inside a room (20), the acoustic field being generated by a sound reproduction system (11) comprising a plurality of loudspeakers (12) and a first computer, the control method being implemented by a control system (10) connected to the plurality of loudspeakers and comprising a plurality of microphones (16) positioned in the listening area and a second computer (18), the method being characterized in that it comprises the steps consisting of, in a calibration phase (101) carried out by the second computer: - the room (20) defining a cut-off frequency between a pressurization regime and a propagation regime, defining (111), in the operating frequency range of the sound reproduction system,at least a first frequency band below the cut-off frequency and a second frequency band above the cut-off frequency; - defining a first filter optimization strategy on the first frequency band dedicated to the pressurization regime by choosing a first target to be reached; - defining a second filter optimization strategy on the second frequency band dedicated to the propagation regime by choosing a second target to be reached; - acquiring (113) a response matrix (^^) of the loudspeakers and the room by applying a test signal to each loudspeaker successively and by measuring an elementary response for the loudspeaker activated at each of the microphone positioning points; and, - determining a global filter as a function of the response matrix,of the first strategy and the second strategy by optimizing the acoustic field in the listening area relative to the first and second targets.

2. MIMO control method according to claim 1, wherein determining a global filter (^^) firstly consists of determining by optimization a first filter (^^, (1) ) for the first band from the response matrix (^^) and the first target and a second filter (^^ (2) ) for the second band from the response matrix (^^) and the second target, then combining the first and second filters to obtain the overall filter.

3. MIMO control method according to claim 1, wherein determining a global filter (^^) firstly consists in combining (216) the first and second targets to obtain a global target (^^), then in determining (217) by optimization the global filter from the response matrix (^^) and the global target (^^).

4. MIMO control method according to claim 1, wherein determining a global filter (^^) consists in: - filtering (316) the response matrix (^^) on the first frequency band, respectively on the second frequency band, to obtain a first response matrix (^^ (1) ), respectively a second response matrix (^^ (2) ) ; - determine (117) by optimization a first filter (^^ (1) ) for the first band from the first response matrix and the first target and a second filter (^^ (2)) for the second band from the second response matrix and the second target; then, - combining (319) the first and second filters to obtain the global filter.

5. MIMO control method according to claim 1, in which a global target (^^) varying continuously with the frequency is determined by interpolation of the first and second targets, then the global filter (^^) is determined by optimization from the response matrix (^^) and the global target.

6. Control method according to any one of claims 2 to 5, in which the first target ^^ ^ ( ^ 1) at a point ^^ of positioning of a microphone is synthesized: - sous une seconde forme : ^^(1) ^^ = ^^0^^(^^ − ^^0), ∀^^ où ^^0 est l’amplitude du acoustic field and ^^0 is a reference time of emission of the test signal; - sous une première forme : ^^(1) ^^ = ^^^^^^(^^ − ^^0), ∀^^ où ^^^^ est l’amplitude du acoustic field and ^^0 is a reference time of emission of the test signal - sous une troisième forme : ℎ^^^^^^^^^^ est la réponse du s-ième haut-parleurmeasured at a single reference point, ^^ ^^^^^^ ; - sous une quatrième forme : ^^(1) ^^ = où ^^^^ est un sous- set of measuring points comprising ^^ ^^ points, and ℎ ^^^^ is the response of the s-th loudspeaker measured at point ^^; or, - sous une cinquième forme : ^^(1) ^^ = ℎ^^^^ .

7. Control method according to any one of claims 2 to 6, in which the second target ^^ ^ ( ^ 2) at a point ^^ of positioning of a microphone is synthesized: - sous une première forme : ^^^ ( ^ 2) = ^^^^^^(^^ − ^^^^), où ^^ est la fonction de Dirac, ^^^^ est l’amplitude de l’onde au point de mesure ^^, et ^^^^ est le retard qui dépend of the measuring position relative to the or each loudspeaker emitting the test signal; - sous une seconde forme : ^^(^^) es une fenêtre temporal applied to the responses ℎ ^^^^ ( ^^ ) measured, and ^^ ^^ is the subset of speakers involved in creating the target, - sous une troisième forme : ou, - sous une quatrième forme , où ^^ est la distance entrespeaker of index ^^ and the measurement point ^^.

8. Control system (11) characterized in that the control system is adapted to implement, when associated with a sound reproduction system, a MIMO control method according to any one of the preceding claims.

9. Control system according to claim 8, in which the control system comprises a first computer and a plurality of microphones and the sound reproduction system comprises a second computer and a plurality of loudspeakers, the first and second computers being the same computer suitably programmed or two separate computers but connected for the implementation of the MIMO control method. 10.Computer program (36) comprising software instructions which, when executed by the computer of a control system according to claim 8 or claim 9 associated with a sound reproduction system, implement a MIMO control method according to any one of claims 1 to 7.

Images