RS20060611A - Technique and system for automatic gain control (agc) using microphone array - Google Patents

Abstract

Invention belongs to the procedure and system for automatic control of a speech signal gain (AGC) recorded by means of a microphone sequence in specific ‘hands-free’ teleconference conditions of communication and in accoustic ambience with large level of various interferences such as: accoustic echo, interferential noise, reverberation and other speakers, as well as a different position of a speaker in space in comparison to the microphone sequence. The procedure for AGC consists from: estimation of trajectory of peak speech power of the speech signal, estimation of powers of accoustic interferences and participation thereof in determination of coeficient of increase, determination of the downward(11)compression characteristics, determination of coefficient of amplification for AGC and formation of speech signal output with minimal distortions. The system consists of one input and one output for the current speech signal, several inputs whereto the signals are brought for accoustic disturbance estimation, and three inputs whereto are set three constants that define the field of active functioning AGC: nominal power of output signal, dynamics of change of automatic gain control (AGC) system and maximal downward slant of compression characteristics. AEC SD VF Power Estimation of Diffused Noise Echo Power Estimation Formula u cetvrtom redu odozgo Slope estimation Gain coefficientPeak Power Estimation Moment Power estimation

Description

PROCEDURE AND SYSTEM FOR AUTOMATIC^ GAIN CONTROL (AGC) BASED ON %|

MICROPHONE ARRAY READINGS

TECHNICAL FIELDS TO WHICH THE INVENTION RELETS

The invention belongs to the field of acoustic signal processing, or more specifically, methods of automatic gain control (AGC) of a speech signal recorded using a microphone array in specific "hands-free" teleconference communication conditions.

TECHNICAL PROBLEM

Hands-free, full-duplex voice communication systems are used in many applications such as: video-telephone systems, teleconferencing systems, speakerphones in rooms or cars, human-computer voice communication, etc. "Hands-free" voice communication implies that the speaker is in an acoustic environment at a certain distance from the interface elements of the system - microphone and speaker. Such conditions of speech communication generate more technical problems that need to be solved in order to maintain the quality of communication at an acceptable level. One of the problems is maintaining the level of the transmitted speech signal at an approximately constant level regardless of the type of speaker or his distance from the microphone.

Namely, three facts are key in the previously indicated problem. First, the intensity of different speakers' voices depends on their phonetic capacity. It is a natural individual characteristic of every person, but it can also be an acquired characteristic depending on social or habitual factors. For example, a speaker from rural areas usually has a higher intensity of speech compared to a speaker from an urban area. Second, when using "hands-free" communication systems, the speaker may be at different distances from the microphone. It is known that the sound intensity decreases with the square of the distance, which in the previously mentioned case introduces significant variations in the size of the speech signal at the microphone output. And thirdly, especially in teleconference applications, the speaker is often in motion and changes his location in space in relation to the position of the microphone, which causes the intensity of the speech signal at the microphone output to vary.

In addition to the mentioned technical problems, there is also the problem of varying the voice intensity in cases where ambient noise, reverberation or other speakers are present in the room, which is a typical case of the "cocktail-party" effect, due to which the current speaker raises the level of his voice (the so-called Lombard effect). All the mentioned technical problems are solved by a technical solution known as "automatic gain control" -AGC (Automatic Gain Control, English). constant its level at its output. When it comes to the speech signal, this control is performed in both domains: amplitude and time.

Two additional problems AGC as a technical solution must solve in the case of teleconferencing applications, namely: a) speech signal discrimination - ambient interference, where reverberation signals and competing speakers should be included, and b) the speed of AGC's reaction to the appearance or disappearance of a speech signal in the analyzed signal, which requires good compatibility of AGC's work with the nature of the speech signal.

The solution of all the mentioned problems should aim at ensuring maximum intelligibility, naturalness and pleasantness of listening to the voice signal in complex teleconference conditions of application of AGSolutions.

STATE OF THE ART

There are two general techniques in the realization of the AGC problem. In the first, the input signal is used to generate a gain coefficient that is multiplied by the input signal and thus produces an output signal of approximately constant power. This technique is called the "feed back" technique. Another technique is the so-called "feedback" technique in which the output signal is used to generate the gain coefficient, which is compared with the reference signal (which defines the desired level of the output signal) and the error signal defines the gain coefficient.

Conventional AGC systems, based on either of these two techniques, solve the issue of automatic voice signal level control based on voice signal amplitude compression using non-linear estimation of the voice signal power and elements of the speech activity detector (VAD-voice activity detector) or the so-called squelch system. For example: U.S. patent 4,947,133, filed Jan. 19, 1988, entitled "Method and apparatus for automatic signal level adjustment," uses a "feedback" technique in adaptive speech signal compression, then U.S. Pat. patent 5,854,845, filed Feb. 21, 1996, entitled "Method and circuit for voice automatic gain control," provides a method for automatically controlling signal gain while minimizing distortion at speech-pause boundaries, as well as U.S. Pat. patent 6,959,082, filed July 10, 2001, entitled "Method and system for automatic gain control with adaptive table lookup", which uses both "feedback" and "feedback" techniques in automatic gain control where the gain is tabulated based on estimated input signal strength.

Somewhat different requirements in the implementation of AGC arise in specific "hands-free" teleconference conditions of full duplex communication. High-quality speech recording in the presence of acoustic disturbances and room reverberation is a complex problem. In conditions where the spectrum of the useful speech signal overlaps with the spectrum of the present disturbances, it is not possible to achieve a significant improvement in the quality of the speech signal with single-channel signal processing procedures, but multi-microphone procedures (microphone arrays) are used. The advantage of microphone arrays over single-channel processing methods is their ability to adapt their spatial reception characteristic (directivity characteristic) to the current spatial arrangement of the selected speaker and interference. At the same time, they achieve the maximum suppression of the present disturbances while at the same time highlighting the selected speaker. The basic problems encountered in the application of microphone arrays are the following (M.S. Brandstein, D.B. Ward (Eds.), Microphone Arrays: Signal Processing Techniques and Applications, Springer, Berlin 2001; Y. Huang, J. Benestv, Audio signal processing for next generation multimedia communication systems, Kluwer Academic Publishers Publ., 2004.): not knowing the exact location of the selected speaker, not knowing the number and spatial arrangement of the interferences present, multiple reflections of the useful source and interference on the walls of the room and the non-stationarity of the source of acoustic interference and the selected speaker.

When the microphone array is used in teleconferencing systems that operate in full duplex, then the number of problems increases. The biggest problem is the appearance of an acoustic echo, followed by double speech activity (activity in both directions of communication), as well as the possible appearance of system instability, the so-called. microphonics. The functioning of AGC is significantly more complicated in these conditions (K. Kobavashi, Y. Haneda, K. Furuva, A. Kataoka, "A hands-free unit with adaptive microphone array for directional AGC", 2005 International Workshop on Acoustic Echo and Noise Control, Eindhoven, The Netherlands, September 12-15, 2005).

The integral solution AGC, presented in this patent, combines the positive features of some conventional solutions with new solutions in a multi-microphone environment and the requirements imposed by the "hands-free" teleconferencing application in full duplex, ensuring quality voice communication.

DISCLOSURE OF THE ESSENCE OF THE INVENTION

The subject of this invention is a method and system for automatic gain control (AGC) of a speech signal recorded using a microphone array in a complex acoustic environment and in specific "hands-free" teleconference communication conditions, with the aim of ensuring the quality and intelligibility of the speech signal.

The essence of the invention lies in the specific processing of the speech signal that is recorded in the acoustic environment of the room where the system and the speaker are located. A microphone array of N microphones is used to record the speaker in the room, located at a certain distance (up to several meters). The microphone array records all the signals in the room: the useful signal as a direct wave arriving from the speaker to the microphone and interference signals that can be varied. The following interference signals appear: acoustic echo as a direct sound wave from the speakers through which the interlocutor's voice is emitted from the far end of the communication channel, direct waves from one or more noise sources or sources of other disturbances that can be found in the room and all reflected waves (room echoes) originating from all sound sources, including the speaker, which are caused by room reverberation. It should be emphasized that the sources of sounds in the room can be stationary or non-stationary, which is the most common case, both according to their characteristics and their location in the room (moving sources of sounds). The essence of the invention is to preserve the quality and intelligibility of the current speaker's speech signal despite large ambient disturbances, different possible spatial positions of the speaker in relation to the microphone array and, consequently, large variations in the intensity of the speech signal.

A specific aspect of the invention is found in the complete signal processing in the frequency domain, which enables certain advantages in terms of processing speed and the number of computational operations, which are very important for real-time implementation.

The next specificity of the invention is the nonlinear characteristic of amplitude compression. The slope of this characteristic is determined in a complex way through the trajectory of the peak power of the speech signal. The analysis includes the trend and convexity of the trajectory, and the relative comparison of these parameters with the current peak power of the speech signal. Therefore, the characteristics of the speech signal are fully taken into account, and the compression characteristic is optimized for this type of signal.

Determination of the gain coefficient is a further specificity of the invention. It is determined on the basis of the slope of the compression characteristic as well as on the basis of the magnitude of the peak power of the input speech signal and the average power of the disturbances that appear in the system of free speech communication based on the microphone array.

The next specificity of the invention is the procedure of adaptive determination of the amplification coefficient based on detected information about pauses in the speech signal, the presence of residual echo in the system and the presence of a competing speaker or acoustic disturbance. This information enables the proper functioning of the AGC with different input signal content and ensures the improvement of the quality of the output signal.

A particular specificity of the invention is a set of parameters that enables the optimal selection of AGC operating conditions in accordance with the application requirements. These parameters define the nominal level of the output signal, the maximum slope of the compression characteristic and the maximum gain of the AGC. In this way, the flexibility of the solution was achieved.

Inventiveness in this invention is found in the improvement of each of the mentioned specificities, and especially in the fact that information about ambient disturbances as well as information about current speech activity are indirectly entered into the amplification coefficient estimation algorithm.

These and other aspects, specificities and benefits of the present invention will be more apparent upon review of the detailed description of the invention, patent claims and accompanying drawings.

BRIEF DESCRIPTION OF THE IMAGES AND DRAWINGS

Figure 1 - shows the ambient conditions of application of the system for free video-telephone communication using a microphone array.

Figure 2 - shows the block diagram of the system for processing audio signals within the system for free video-telephone communication and locating the module for AGC; the basic blocks of this system are: the echo cancellation block (AEC), the microphone array directivity characteristic formation block (SD-BF), the speaker localization block (DOA), the noise cancellation block (NR) and the automatic gain control (AGC) block.

Figure 3 - shows the process of forming the transmission characteristic of AGC.

Figure 4 - shows a block diagram of the basic structure of the automatic gain control (AGC) system.

Figure 5 - shows the flow of operations in the calculation of the amplification factorAagc.

Figure 6 - shows the interface points and signals of the AGC system.

DETAILED DESCRIPTION OF THE INVENTION

This invention describes a system and method for automatic gain control (AGC) of a speech signal in a free speech communication system using a microphone array. In order to understand the need and importance of AGC in such systems, Figure 1 shows the ambient conditions for free speech communication, while Figure 2 shows the block diagram of the audio signal processing system in which the voice signal AGC module is located.

Figure 1 schematically shows the ambient conditions of application of the system for free video-telephone communication using a microphone array. In the room 101 there is a system for free video-telephone communication, a speaker 111 and a noise source 114, which is common for any acoustic environment. Through the speaker 102 of the stereo audio system, the speaker 111 listens to the incoming speech signal 104 of the interlocutor from the far end, usually as a mono signal; it is possible that another audio signal is transmitted through the stereo audio system. Sound in the room environment 101 is recorded by a microphone array 103 composed of N microphones. After the complex processing of microphone signals in block 107, the speech signal of the speaker 111 is transmitted via block 108 to the remote interlocutor as a mono signal.

The ambient conditions for speech communication in room 101 are very complex. During free video-telephone communication, there are at least three sound sources in the room 101: stereo speakers 102 that emit the speech of the remote interlocutor and another audio signal, speaker 111 and at least one source of noise 114. There may be several sources of noise in the room: computer noise, air conditioning system noise, noise from the street penetrating the room through the windows, noise from neighboring rooms, building vibrations, or another speaker, multiple speakers, music source, etc. Thus, a very complex acoustic picture appears in the room. The microphone array 103 records, as a sensor system, all sounds in the room, it records direct sound waves from each source, but also all reflections from the walls of the room and other objects in it. So, for example, from the speaker 102 to the microphone array 103, a direct wave 109 and many reflected waves arrive, of which only one 110 is shown in Figure 1; from the speaker 111 comes a direct wave 112 and, in addition to the others, two reflected waves 113a and 113b, from the noise source 114 a direct wave 115 arrives and, in addition to the others, a reflected wave 116.

Of all the sounds that the microphone array records, only the direct wave 112 from the speaker 111 is a useful signal, all others are interference. The acoustic echo 109 coming from the speaker 102 is usually the most intense disturbance. All other reflections collectively make up the reverberation of the room. The task of the audio signal processing block 107 is to suppress the acoustic echo signal, to select the useful signal 112 from all other interferences, to suppress the reverberation signals and to suppress the direct signals of the interference sources, which may be more than one source. The special task of block 107 is to adaptively monitor the non-stationarity of the acoustic scene in the room, whether the speaker is moving, or is in different positions in the room from conversation to conversation, or that noise sources are moving, are non-stationary or change their characteristics. From the point of view of the remote interlocutor, the outgoing voice signal 108 should be stable and independent of the aforementioned acoustic variability, which should ensure quality and pleasant voice communication.

Figure 2 shows a block diagram of the system for processing audio signals within the system for free video-telephone communication, which is marked 107 in Figure 1. Microphone signals 103, from Ml to MN, as well as stereo speaker signals 102, Zv-L and Zv-D, are introduced via the input interface 201 to block 202, N-channel acoustic echo canceller - AEC (Acoustic Echo Cancelling), in which suppresses the acoustic echo 109, Fig. 1, in the recorded speech signal 112, Fig. 1. In the input interface 201, the signal is converted into the frequency domain using the Discrete Fourier Transform (DFT) and all further processing of the signal is performed in the frequency domain, at the level of the block of input measurements. At the output of block 202, signals are received from the microphone with suppressed acoustic echo, SaecidosAecn- These signals are introduced into block 203, the superdirective beamformer - SD-BF (Superdirective Beamformer), which shapes the directionality characteristic of the microphone array 103, and into block 204, the azimuth DOA (Direction Of Arrival), which determines the location of the current speaker in the horizontal plane and provides information about the angle azimuth6a. This data is used to direct the characteristic of the microphone array towards the current speaker.

The signal at the output of block 203 contains the useful speech signal and the interference signal consisting of the residual signal after acoustic echo suppression, suppressed ambient noise and suppressed reverberation signals. This signal enters block 205, the noise suppressor - NR (Noise Reduction), where additional interference signal suppression is performed. The suppression process is adaptive considering the non-stationarity of the interference signal.

The final block of signal processing in the system for free speech communication is block 206, automatic gain control - AGC, which is the subject of this invention. In this block, more information from the entire system is used, which is important for defining the possible conditions in which the speech signal can be found and where it is necessary to perform its amplitude correction in an appropriate manner. In this way, it is possible to provide approximately the same level of the transmitted speech signal regardless of the distance of the current speaker from the microphone array and ensure its better quality at the far end of the communication channel. Through the output interface 207, the estimated speech signal at the near end is transmitted through the communication channel to the remote interlocutor.

It is important to note that the functioning of AGC in complex ambient conditions of microphone array application, in multi-channel systems and in free speech communication, is very different from the conventional application of AGC in single-channel systems and applications.

Figure 3 shows the process of forming the transmission characteristic of the AGC system. The diagrams are given in the form Low, = /(L,„). In the case of a system without an AGC function with unity gain, the transfer characteristic would be 301,Ki, which provides only input to output forwarding. The basic task of the AGC system is to maintain the speech signal at its output at the same optimal power level 302,Lopl, and in the ideal case - when we have full compression, each input level would be brought to the Lopt level.

The transmission characteristic K2, which achieves complete compression of the input dynamics, is not good in conditions of the presence of noise in the system, because in that case the noise is amplified and equalized in level with the speech signal. A compromise solution is to use the characteristic 304,K3 instead of the transmission characteristic 303,K2, which reduces (compresses) the input dynamics according to the given compression constant. The characteristic 304,K3 is described in the decibel diagram by a linear function:

and where: Pm- signal power at the input to the compressor, Pout- signal power at the output from the compressor, Pn0m- nominal output power to which the output signal is to be set and Po - reference power in relation to which the levels are measured.

When the input signal is Lopttada the level at the output is also Lopta linear amplification Aagcza that is achieved by this process should be 1. If:

then for amplification A, inserting (2) and (3) into (1), we get: mog( AlgcPJP0)=lOa\ og( Pin/ P0)+ (1 -a)101og(P„om /P0), (4)

namely:

From (5) we can see the following property of amplification Aagc. When the power of the input signal Pjnto zero, with the condition 0<ot<l, then the amplification Aagc tends to infinity. This feature is not desirable, because in this case there is a large and unnecessary amplification of very weak signals, which most often represent noise. Therefore, the relation (5) is modified by adding the term 5<P> to the denominator, and we get here:

For extremely weak signals, relation (6) limits the gain to the value:

Also, by relation (6), the transfer characteristic 304,K3 is modified into the characteristic 305,K4 (Figure 4). Three characteristic zones can be observed on the characteristic K4. Zone 1, when the level of the input signal is much lower than the value SPnom. In that case, the gain Aagc is constant according to relation (7). It is obvious that the input signal should not be in that zone, because there the gain is constant and there is no effect of AGC. Zone 2 represents the optimal working area. For weak signals, the amplification is limited to Aagcmax, while the compression of signal dynamics defined by the constant a is applied to the part around Lop. In zone 3, there is a constant degree of compression of the input signal, which is defined by the compression constant a.

In order to bring the input signal to one of zones 1, 2 or 3, the input signal should be amplified before the AGC block with a fixed gain, which moves the characteristic 301,Ki, figure 3, to the position 306,K' i.

Figure 4 shows the block diagram of the AGC system. Its task is: (1) to amplify weak speech signals and to weaken excessively strong signals according to a predetermined characteristic of compression of signal dynamics, (2) to reduce the gain on parts of the input signal where there is only an echo of the signal, stationary noise or competing speech-interference, in order to silence these interferences sufficiently and (3) to silence parts of the input signal where both a useful speech signal and interference are simultaneously present, while preserving speech intelligibility.

The input signal sin, which is equal to the signal of block 205 from Figure 2, is multiplied in block 401 with the amplification of Aagciz of block 402, and thus the output signalout-sagc-AmplificationAagcje according to the relation (6) is determined by four parameters: Pnom- is a fixed parameter that defines the desired level of the output signalasout=sAgc, 8 is a parameter that limits the amplification of the input signalasin = sNR valueA„ gcmax (for example: for S=0.001 the maximum possible amplification is Aagcmax = 31.6, equation (7)), P[ n - the signal power at the input to the compressor and a-compression constant that defines the slope of the compression characteristic (Figure 3). input power estimation signalPin, which contains the useful signal of the current speaker and all interference signals. With this second shift, the functional dependence of the Aagcod interference signal was introduced, which proved to be very useful from the aspect of preserving the quality of the output speech signal. Now relation (6) becomes:

The estimation of the slope of the compression characteristic is performed in block 403, while the estimation of the input power P is performed in block 404. Both of these blocks require an estimation of the peak power of the useful speech signal Pd which is determined as follows. First, in block 405, the estimation of the current power of the current speaker from the selected direction of the microphone array is determined, according to Figure 2, as follows: where: C' f = [c, f c2 f ■ ■ ■ cN f ]weighting coefficients of the superdirective spatial filter SD-BF,X = [ sAEa sAEC2 •••<s>AECN]',/min<i>fimx are the minimum and maximum frequency in the DFT spectrum of the analyzed speech signals, respectively. Then, in block 406, the peak power of the signal from the selected direction Pd (t) is evaluated as follows:

where: n- the current block of signal processing and ic?- a constant close to 1.

In block 404, as information, the data, figure 4, about the estimation of the current power of the current speaker Pd (relation (9)) enters. Another piece of information is the estimate of the power of the unsuppressed echo Peh0, which is determined as follows:

where:y is the estimated echo signal obtained in block 202, Figure 2, while isecho is the echo suppression constant.

The third piece of information that enters block 404 is a rough estimate of the diffuse noise power Pn. This power is obtained by processing the output signals of block 202, block AEC of Figure 2, and the input signal of block AGC. Namely, the signalisAEcidosAEcnne contain the acoustic echo that is suppressed in block 202, that is, they contain the current speech signal and all other interference signals. If the estimated power of the current speaker is taken from the average power of the signal, a rough estimate of the power of all other disturbances in the recorded environment, primarily diffuse noise, is obtained. This can be expressed by the following relation: where is the average power at the output of the AEC block and it is determined as follows:

Therefore, the output from block 404 is a power estimate, which, in addition to the peak power of the speech signal, also contains estimates of the interference power in the recording environment according to the relation:

In block 403, the slope of the compression characteristic is estimated and the slope is calculated. The calculation procedure is shown in Figure 5.

By analyzing the trajectory of the peak power of the speech signal, two parameters are calculated: the growth trend, in block 501, and the convexity of the power trajectory, in block 502, in order to make a "soft decision" about the presence of speech in the input signal, according to the relations: convexity—max { convexity_,0 } (17) Then, the complex of variables (n) is calculated, in block 503, with the following procedure:

Finally, the magnitude slope, which represents the degree of compression of the signal dynamics and enters relation (8), is calculated in block 505 on the basis of the previously calculated magnitude gamma_ and the magnitude slopemax, which represents the given maximum value of the slope.

The magnitude of the slope is calculated by the relation

The limit values of the slope_size are:

Thus, all quantities are determined by the equation (8) for calculating the gain Aagc.

Now we can see the role of size in the calculation of Aagc, equation (8). If the power level of the residual echo signal or the level of diffuse ambient noise, which cannot be canceled but only reduced, is higher, a lower amplification is realized by means of the AGC control compared to the case where there would be no interference. On the other hand, if in the input signal s,„ =s^ there is no speech signal but only residual noise after the noise reduction in the NR block, block 205 of Figure 2, then P^ tends to the minimum value as well as all quantities in (15) to (21), which causes a decrease in Aagcna minimum value, that is, A—»1.

This invention describes the process of processing acoustic and speech signals in a free speech communication system that functions in full duplex, and which refers to the automatic control of the output level of the speech signal. The specificity of this invention is its application in systems based on microphone arrays, its complete realization in the frequency domain and the possibility of its application in other communication systems such as video-telephone systems, teleconference systems, speakerphones in rooms or cars, human-computer communication by voice, etc.

The procedures and techniques for processing acoustic and speech signals in this invention are independent of the number of microphones in the microphone array, Figure 2, and are under the control of a number of parameters that enable the optimization of solutions for different applications.

The acoustic and speech signal processing methods and techniques of the present invention can be implemented in a variety of ways. For example, these techniques can be implemented in hardware, software, or a combination. A hardware implementation may use specific integrated circuits (ASICs), digital signal processors (DSPs), programmable logic circuits (PLDs or FPGAs), and other electronic circuits designed to perform the functions described in this invention. In a software implementation, program codes can be stored in memory units and executed by a processor such as a PC, PDA, DSP, etc.

Figure 6 shows the interfaces of the AGC system as an independent hardware/software solution 600. In addition to the input 601 and output 602, for the input speech signal and the output processed speech signal, this AGC solution also has three control inputs 603, 604 and 605, through which the parameters are introduced into the system, Pnomi8, respectively, which define the optimal operation of the AGC system. In addition, the AGC system contains (N + 1) inputs as an interface with a system based on a microphone array of N microphones, intended for hands-free voice communication in specific teleconferencing application conditions.

The details of the invention described herein enable any person skilled in the art to implement the generic principles of the invention in other speech communication systems without departing from the scope of the invention.

Claims (18)

1. A procedure for automatic gain control (AGC) based on the reading of the microphone array, characterized by: adaptive determination of the gain coefficient, which directly controls automatic control of the amplification of the output signal and maintains its level within defined limits; adaptive estimation of the slope of the compression characteristic by which it is compressed input signal; adaptive estimation of the peak power of the input signal, which is the basic input parameter for estimating the slope of the compression characteristic and the amplification coefficient; adaptive estimation of interference signal strength based on the readings of the microphone array in system of free speech communication. 2. The procedure according to claim 1, characterized by the fact that the complete processing of all audio signals is performed in the frequency domain. 3. The method according to claim 1, characterized by the fact that the coefficient of amplification of the input signal is determined adaptively depending on the power of the input speech signal and the power of the interference signal. 4. The procedure according to claim 3, characterized by the fact that the input signal amplification coefficient is controlled by two constants: the nominal output power to which the output signal is to be adjusted and the coefficient that limits the maximum amplification of the AGC system. 5. The method according to claim 3, characterized by the fact that the amplification coefficient of the input signal is determined based on the adaptive estimation of the slope parameter of the compression characteristic by means of which the input signal is compressed. 6. The procedure according to claim 1, characterized by the fact that the slope coefficient of the compression characteristic is estimated on the basis of: two predefined constants that limit the zone of action of AGC and whose by choosing, the AGC function can be optimized in various applications, and based on the adaptive estimation of the peak power of the input signal, which directly determines the current value of the gain coefficient. 7. The procedure according to claim 6, characterized by the fact that the slope coefficient of the compression characteristic is estimated on the basis of two predefined constants: the nominal output power to which the output signal is to be set and the coefficient that defines the maximum allowed value of the slope of the compression characteristic. 8. The method according to claim 6, characterized by the fact that the slope coefficient of the compression characteristic is adaptively estimated based on the adaptive estimation of the peak power of the input signal and the determination of its trajectory in time. 9. The method according to claim 8, characterized by the fact that the slope coefficient of the compression characteristic is determined based on the size of the growth trend of the trajectory of the peak power of the input signal, adaptively calculated on several previous blocks of analysis of the peak power of the input signal. 10. The method according to claim 8, characterized by the fact that the slope coefficient of the compression characteristic is determined based on the convexity of the trajectory of the peak power of the input signal, adaptively calculated on several previous blocks of analysis of the peak power of the input signal. 11. The method according to claim 8, characterized by the fact that the slope coefficient of the compression characteristic is determined on the basis of a "soft decision" about the presence of speech in the input signal using the parameters of the growth trend and the convexity of the trajectory of the peak power of the input signal. 12. The method according to claim 1, characterized by the fact that the estimation of the peak power of the input signal is determined adaptively in such a way that when the power increases, the current power estimation is adopted, while when the power decreases, recursion with a certain time constant is used. 13. The procedure according to claim 1, characterized by the estimation of the strength of the unsuppressed echo, based on the estimation of the strength of the echo signal in the part of the free speech communication system that performs echo suppression and the application of the estimated echo suppression factor to the input to the AGC, which is required to determine the AGC amplification coefficient. 14. The method according to claim 1, characterized by the fact that a rough estimate of the power of the diffuse noise is made, in the acoustic environment in which the microphone array records, based on the difference between the estimated mean power of the microphone signals, in which echo suppression was performed, and the estimated current power of the current speech signal, which is needed to determine the AGC amplification coefficient. 15. A system for automatic gain control (AGC) based on the reading of the microphone array, characterized by: a block for adaptive determination of the gain coefficient, which directly controls automatic control of the amplification of the output signal and maintains its level within defined limits; block for adaptive estimation of the slope of the compression characteristics by means of which compresses the input signal; block for adaptive estimation of the peak power of the input signal, which is the basic input parameter for estimating the slope of the compression characteristic and the amplification coefficient; block for adaptive estimation of interference signal strengths based on readings of the microphone array in the system of free speech communication. 16. The system according to claim 15, characterized by the fact that its functioning is optimized for use in free speech communication systems based on microphone arrays in which acoustic echo suppression (AEC) and locating the speaker in space based on the directional characteristic (BF) of the microphone array are included, whereby the number of microphones in the array is not a limiting factor. 17. The system according to claim 15, characterized by the fact that it contains one input and one output for the current speech signal and several inputs to which signals from free speech communication systems are fed and which enable the estimation of all acoustic disturbances in the environment of recording the microphone array. 18. The system according to claim 15, characterized by the fact that it contains three inputs where three constants are set that define the field of active operation of the AGC, namely: the nominal power of the output signal, the dynamics of the gain change of the AGC system and the maximum slope of the compression characteristic.