MXPA99000022A - System and method to cancel the interference of adaptac - Google Patents

Abstract

A system and method of adaptation is proposed to reduce the interference in a signal received from a set of sensors. Adapter filters are used to generate cancellation signals that closely approximate the interference present in the received signal. The weights of the adaptation filters are converted into the frequency domain where the frequency representation values in a selected frequency range are truncated to avoid signal leakage involving narrowband signals. Discoloration filters are used to produce cancellation signals that have a flat frequency spectrum. A normalized power difference is used to limit the operation of the adaptation filters in the case in which a certain direction interference must be eliminated.

Description

"SYSTEM AND METHOD TO CANCEL THE ADAPTATION INTERFERENCE" BACKGROUND OF THE INVENTION The present invention relates generally to signal processing, and more specifically to a system and method of processing adaptive signals to reduce interference in a received signal. There are many cases where it is desirable to have a sensor capable of receiving an information signal from a specific signal source wherein the environment includes sources of interference signals at locations other than that of the signal source. One of these cases is the use of microphones to record a voice of the specific part in a room where there are other parties speaking simultaneously, which cause interference in the received signals. If the exact characteristics of the interference are known, a fixed weight filter can be used to suppress it. It is often difficult to predict the exact characteristics of the interference, measures that may vary according to changes in sources of interference, background noise, the acoustic environment, the orientation of the sensor with respect to the source of signals, the transmission trajectories from the signal source to the sensor and many other factors. Therefore, in order to suppress this interference, an adaptation system is needed that can change its own parameters in response to a changing environment. An adaptation filter is an adaptation system that can change its own filtering characteristics in order to produce a desired response. Typically, the filter weights defining the characteristics of an adaptation filter are continuously updated so that the difference between the signal representing a desired response and an output signal of the adaptation filter is minimized. The use of adaptation filter to reduce interference in a received signal will be known in the art as adaptation noise cancellation. It is based on the idea of canceling a noise component of a signal received from the direction of a signal source, sampling the noise independently of the source signal and modifying the mixed noise to approximate the noise component in the received signal using an adaptation filter. For a seminal article, in the adaptation noise cancellation, see B. Idrow et al., Adaptive Noise Canceling: Principies and Applications, Proc. IEEE 63: 1692-1716, 1975. A basic configuration for canceling adaptive noise has been a primary input received by a microphone addressed to a desired signal source and a reference input received independently by another microphone directed to a noise source . The primary input contains both a component of the source signal that originates from the signal source and a noise component that originates from the noise source. The noise component is different from the reference input representing the noise source itself, because the noise signal must go from the noise source to the signal source in order to be included in the noise component. The noise component, however, will probably have some correlation with the reference input because both originate from the same source of noise. Therefore, a filter can be used to filter the reference input to generate a cancellation signal that approximates the noise component. The adaptation filter does this by dynamically generating an output signal that is the difference between the primary input and the cancellation signal, and adjusting its filter weights to minimize the quadratic average value of the output signal. When the filter weights settle, the output signal effectively doubles the source signal essentially free of the noise component because the cancellation signal closely follows the track of the noise component. The adaptation noise cancellation can be combined with beam formation, a technique known to use a sensor array to improve the reception of signals coming from a specific direction. A beamformer is a special filter that generates a single channel of the multiple channels received through the multiple sensors, filtering the individual multiple channels and combining them in such a way as to extract the signals coming from a specific direction. Therefore, a beamformer can change the direction of the receiving sensitivity without physically moving the sensor array. For details on beam formation, see B.D. Van Veen and K. M. Buckley, Beamforming: A Versatile Approach to Spatial Filtering, IEEE ASSP Mag. 5 (2), 4-24. Since the beam former can effectively be oriented in many directions without physically moving its sensors, the beam former can be combined with adaptive noise cancellation to form an adaptive beam former that can suppress specific directional interference instead of general background noise. The beam former can provide the primary input by especially filtering the input signals from a sensor array so that its output represents a signal received in the direction of a signal source. Similarly, the beamformer can provide the reference input by spatially filtering the sensor signal so that the output represents a signal received in the direction of the interference sources. For a seminal article on adapters of the adaptation beam see L.J. Griffiths & C.W. Jim, An Alternative Approach to Linearly Constrained Adaptive Beamforming, IEEE Trans. Ant. Prop. AP-30: 27-34, 1982. A problem with a conventional adaptation beamformer is that its output characteristics change depending on the input frequencies and the directions of the sensor with respect to the sources of interference. This is due to the sensitivity of a beam former at different input frequencies and sensor directions. A uniform output behavior of a system through all the input sequences of interest and through all directions of the sensor is clearly desirable in a directional microphone system where reliable reproduction of a sound signal is required regardless of where the microphones are placed. Another problem with the adaptive beam former is the "signal leak". The adaptive noise cancellation is based on an assumption that the reference input representing the noise sources is not correlated with the component of the source signal with the primary input, implying that the reference input should not contain the source signal. But this assumption of "free signal" reference input is violated in any environment. Any mismatch in the microphones (amplitude or phase) or its related analog front end, any reverberation caused by the surrounding environment or a mechanical structure, and even any mechanical coupling in the structure of the physical microphone, will result in "sealing leakage" of the source of signals to the reference input. If there is any correlation between the reference input and the source signal component in the primary input, the adaptation process via the adaptation filter causes the cancellation of the component of the source signal, resulting in distortion and degradation in the source signal. the performance. It is also important to limit the process of adaptation to the case where there is at least some directional interference that is going to be eliminated. Since non-directional noise, such as wind noise or vibration noise induced by the mechanical structure of the system, is not typically correlated with the noise component of the received signal, the adaptation filter can not generate a signal from cancellation that approaches the noise component. The prior art suggests inhibiting the adaptation process of an adaptation filter when the signal-to-noise ratio (SNR) is raised, based on the observation that an intense source signal tends to leak towards the reference input. For example, U.S. Patent Number 4,956,867 describes the use of cross-correlation between two sensors, in order to inhibit the adaptation process when the SNR is high. But the prior art approach fails to consider the effect of directional interference because the SNR-based approach considers only non-directional noise. Since the non-directional noise is not correlated with the noise component of the received signal, the adaptation process seeks in vain new filter weights, which frequently results in the cancellation of the component of the signal source of the received signal .

The prior art approach also fails to take into account signal leakage when the source signal is narrow bandwidth. In a directional icrophor application, the source signal often contains a narrow band signal such as the speech signal, with its spectral density of energy concentrated in a narrow frequency range. When a signal leakage occurs due to an intense narrowband signal, the prior art approach may not inhibit the adaptation process because the total signal strength of this narrowband signal may not be high enough. The component of the source signal of the received signal is canceled as a result, and if the source signal is a voice signal, degradation of the speech intelligibility occurs. Therefore, there is a need for an adaptation system that can suppress bidirectional interference in a received signal, with uniform frequency behavior through a wide angular distribution of the interference sources.

COMPENDIUM OF THE INVENTION Accordingly, an object of the present invention is to suppress interference in a received signal by using an adaptation filter to process the inputs of a sensor array. Another object of the invention is to limit the adaptation process of this adaptation filter to the case where at least some bidirectional interference is to be eliminated. A further object of the invention is to control the adaptation process to prevent signal leakage for narrow band signals. Another object is to produce an output with a uniform frequency compartment in all directions from the formation of sensors. These and other objects are achieved in accordance with the present invention, which uses a system to process the digital data that represents the signals received from a sensor array. The system includes a main channel matrix unit for generating a main channel representing signals received in the direction of a signal source where the main channel has a source signal component and an interference signal component. The system includes a channel matrix unit of preference for generating at least one reference channel wherein each reference channel represents signals received at addresses other than that of the signal source. The system uses adaptation filters to generate cancellation signals that approximate the component of the reference signal of the main channel, and a difference unit to generate a digital output signal by subtracting the cancellation signal from the main channel. Each adaptation filter has a weight update means to find new filter weights based on the output signal. The system includes a weight restriction means for truncating the new weight values of the filter to predetermined threshold values, when each new weight value of the filter exceeds the corresponding threshold value. The system may further include at least one fading filter to generate a flat frequency reference channel. The system may further include an inhibiting means for calculating the energy of the main channel of the energy of the reference channels and for generating an inhibition signal to the updating means by weight based on the normalized energy difference between the main channel and the reference channels. The system produces an output essentially free of directional interference with a uniform frequency behavior in all directions from the system. The objects are also achieved in accordance with the present invention, using a method that can be easily implemented in a program that controls a commercially available DSP processor.

BRIEF DESCRIPTION OF THE DRAWINGS The objects, features and advantages of the present invention will become more readily apparent from the following detailed description of the invention, in which: Figure 1 is a functional diagram of a total system; Figure 2 is a functional diagram of a sampling unit; Figure 3 is a functional diagram of an alternative embodiment of a sampling unit; Figure 4 is a schematic illustration of the derived delay lines, used in a main channel matrix and a reference matrix unit; Figure 5 is a schematic illustration of a main channel matrix unit; Figure 6 is a schematic illustration of a reference channel matrix unit; Figure 7 is a schematic illustration of a decolorization filter; Figure 8 is a schematic illustration of an inhibition unit based on directional interference; Figure 9 is a schematic illustration of a frequency selective restriction adaptation filter; Figure 10 is a functional diagram of a weight-selective, frequency-selective unit; Figure 11 is a flow chart illustrating the operation of a program, which can be used to implement the invention.

DETAILED DESCRIPTION OF THE INVENTION Figure 1 is a functional diagram of a system in accordance with a preferred embodiment of the present invention. The illustrated system has a sensor array 1, a sampling unit 2, a main channel matrix unit 3, a reference channel matrix unit 4, a fading filter set 5, an adaptation filter set 6 frequency-selective constraints, a delay 7, a difference unit 8, an inhibition unit 9, and a D / A output unit 10.

- - The sensor array 1, which has individual sensors a Id, receives the signals from a signal source on the system axis and from the inter-reference sources placed outside the system axis. The sensor array is connected to the sampling unit 2 to sample the received signals, having individual sampling elements 2a to 2d, where each element is connected to the corresponding individual sensor to produce the digital signals 11. The outputs of the sampling unit 2 is connected to the main channel matrix unit 3, producing a main channel 12 representing the signals received in the direction of a source. The main channel contains both a source signal component and an interference signal component. The outputs of the sampling unit 2 are also connected to the reference channel matrix unit 4 which generates the reference channels 13 which represent the signals received from addresses other than that of the signal source. In this way, the reference channels represent interference signals. The reference channels are filtered through the decolorization filters 5 which generate flat frequency reference channels 14 having a frequency spectrum whose magnitude is essentially flat across a frequency range of interest. And the flat frequency reference channels 14 are fed to the set of frequency selective restriction matching filters 6, which generate the cancellation signals 15. Meanwhile, the main channel 12 is delayed through the delay 7 so that it is synchronized with the cancellation signals 15. The difference unit 8 then subtracts the cancellation signals 15 from the delayed main channel to generate a digital output signal 16 that is converted by the D / A unit 10 in analog form. The digital output signal 15 is fed back to the adaptation filters to update the filter weights of the adaptation filters. The flat frequency reference channels 14 are fed to the inhibition unit 9, which calculates the energy or power of each flat frequency reference channel as well as the energy of the main channel and generates an inhibition signal 19 to prevent the escape of sign Figure 2 illustrates a preferred embodiment in the sampling unit. A sensor array 21 having sensor elements 21a-21d is connected to an analog front end 22, which has amplifier elements 22a to 21d, wherein each amplifier element is connected to the output of the corresponding sensor element. In a directional microphone application, each sensor can be a microphone either directional or omnidirectional. The analog front end amplifies the received analog sensor signals to match the input requirement of the sampling elements. The analog front end outputs are connected with a set of delta-sigma 23 A / D converters, where each converter samples and digitizes the amplified analog signals. Delta-sigma sampling is a well-known A / D technique using both oversampling and digital sampling. For details on delta-sigma sampling, A / D, see Crystal Semiconductor Corporation, Application Note: Delta-Sigma Techniques, 1989. Figure 3 shows an alternative embodiment of the sampling unit. A sensor array 31, having the sensor elements 31 a to 31 d, is connected to an amplifier 32 having elements of the amplifier 32 a to 32 d, wherein each amplifier element amplifies the signals received from the corresponding sensor element. The outputs of the amplifier are connected to a sampling and holding unit 33 (S / H) having sampling and holding elements 33a to 33d, wherein each S / H element samples the amplified analog signal of the corresponding amplifying element to produce a discrete signal. The outputs of the S / H unit are multiplexed into a single signal through a multiplexer 34. The output of the multiplexer is connected to a conventional A / D converter 35, to produce a digital signal. Figure 4 is a schematic illustration of the derived delay lines used in the main channel matrix unit and the reference channel matrix, in accordance with a preferred embodiment of the present invention. The derived delay line used herein is defined as a non-recursive digital filter, also known in the art as a cross-filter, a finite impulse response filter or an FIR filter. The illustrated embodiment has four derivative delay lines 40a to 40d. Each derived delay line includes delay elements 41, multipliers 42 and adders 43. Digital signals 44a to 44d are fed to the set of derived delay lines 40a to 40d. The delayed signals through the delay elements 41 are multiplied by filter coefficients, F-j, 45 and added to produce the outputs 46a to 46d. The n-th sample of an output of the derivative delay line i-th, Yj_ (n), can then be expressed as: Y-¡(n) =? ^ J = o F ± rj Xj_ (nj), in where k is the filter length and X-_ (n) is the n-th sample of an input to the derived delay line i-th.

Figure 5 illustrates the main channel matrix unit for generating a main channel in accordance with a preferred embodiment of the present invention. The unit has derived delay lines 50a to 50d as an input section that takes inputs 51a to 51d from the sampling unit. Its output section includes multipliers 52a to 52d in which each multiplier is connected to the corresponding derivative delay line and an adder 53 that sum all the output signals of the multipliers. The unit generates a main channel 54 as a weighted sum of the outputs of all the multipliers. The weights of the filter 55a through 55d can be any combination of fractions as long as their sum is 1. For example, if four microphones are used, the mode can use the filter weights of a quarter to take into account the contribution of each microphone. The unit acts as a beamformer, a spatial filter that filters a signal that comes in all directions to produce a signal that comes in a specific direction without physically moving the sensor array. The coefficients of the derived delay lines and filter weights are graded in such a way that the received signals are spatially filtered to maximize sensitivity to the signal source. Since some interference signals find a way to reach the signal source due to many factors such as the reverberation of a room, the main channel 54 representing the signal received in the direction of the signal source contains not only a component of the source signal but also a component of the interference signal. Figure 6 illustrates the reference channel matrix unit for generating reference matrix channels, in accordance with a preferred embodiment of the present invention. It has derivative delay lines 60a to 60d, as a section that takes the inputs 61a to 61d from the sampling unit. The same derivative delay lines as those in Figure 4 can be used, in which case, the derived delay lines can be shared by the main and reference channel matrix units. Its output section includes multipliers 62a to 62d, 63a to 63d, 64a to 64d and adder 65a to 65c, where each multiplier is connected with a corresponding derivative delay line and an adder. The unit acts as a beamformer generating the reference channels 66a to 66c, which represents the signals that - they arrive off the axis from the source of signals, obtaining the weighted differences of certain combinations of output from the derived delay lines. The combinations of filter weight can be any number as long as their sum of filter weights to combine a given reference signal is 0. For example, the illustrated mode can use a combination of filter weights (Wll, W12 , W13, W14) = (0.25, 0.25, 0.25, -0.75), in order to combine the signals 61a to 61d to produce the reference channel 66a. The net effect is to place a null value (low sensitivity) on the receiving gain of the beamformer towards the signal source. As a result, the reference channels represent the interference signals in directions other than that of the signal source. In other words, the unit "directs" the digital input data to obtain interference signals without physically moving the sensor array. Figure 7 is a schematic illustration of the decolorization filter in accordance with a preferred embodiment of the present invention. It is a derivative delay line that includes delay elements 71, multipliers 72 and summing 73. A reference channel 74 is fed into the derived delay line. The delayed signals are multiplied by the filter coefficients, Fj_, 75 and added to produce an output 76. The filter coefficients are graded in such a way that the filter amplifies the low-magnitude frequency components of an input signal, in order to obtain an output signal having an essentially flat frequency spectrum. As mentioned above in the background section, the output of a conventional adaptation beamformer suffers from a non-uniform frequency behavior. This is because the preference channels do not have a flat frequency spectrum. The receiver sensitivity of a beamformer to a specific angular direction is often described in terms of a gain curve. As mentioned above, the reference channel is obtained by placing a null value on the gain curve (making the formation of sensors insensitive) in the direction of the signal source. The resulting gain curve has the lowest gain for the lower frequency signals than the higher frequency signals. Since the reference channel is modified to generate a cancellation signal, a non-planar frequency spectrum of the reference channel is transferred to a non-uniform frequency behavior at the system output.

The fading filter is a fixed coefficient filter that flattens the frequency spectrum of the reference channel ("fading", hence the reference channel) by reinforcing the low frequency portion of the reference channel. By adding the discoloration filters to all the outputs of the reference channel matrix unit, an essentially flat frequency response is obtained in all directions. The fading filter in the illustrated mode uses a derivative delay line filter that is the same as the finite impulse response (FIR) filter, but other filter classes such as an infinite impulse response filter (IIR) can also be used. ) for the discoloration filter, in an alternative mode. Figure 8 schematically illustrates the inhibition unit in accordance with a preferred embodiment of the present invention. It includes energy calculation units 81, 82 that calculate the energy of a main channel 83 and of each reference channel 84, respectively. A sample power calculation unit 85 calculates the energy of each sample. A multiplier 86 multiplies the energy of each sample by a fraction,, which is the reciprocal of the number of samples for a given averaging period in order to obtain an average sample energy 87. An adder 88 adds the average sample energy to the output of the other multiplier 89, which multiplies an average 90 of main channel energy previously calculated by (1-a). A new average of main channel energy is obtained by (new sample energy) x a + (old energy average) x (1-a). For example, if you use a sample average of 100, a = 0.01. The updated energy average will be (new sample energy) x 0.01 + (old energy average) x 0.99. In this way, the updated energy average will be available during each sampling time instead of after a period of averaging. Even though the illustrated majority shows a method of calculating the average energy, other types of energy calculation methods can also be used in the alternative mode. A multiplier 91 multiplies the main channel energy 89 with a threshold 92 to obtain an average 93 power or normalized main channel power. An adder 93 subtracts the reference channel energy averages 95 from the average normalized main channel energy 93 to produce a difference 96. If the difference is positive, a comparator 97 generates an inhibition signal 98. The inhibit signal is provided to the adaptation filters to stop the adaptation process to prevent signal leakage.

Even though the illustrated modality normalizes the main channel energy average, an alternative modality can normalize the average reference channel energy instead of the average main channel energy. For example, if the threshold 92 of the illustrated mode is 0.25, the same effect can be obtained in the alternative mode by normalizing each reference channel energy average by multiplying it by 4. This inhibition approach is different from the inhibition approach based on SNR of the prior art, mentioned in the background section since it detects the presence of significant directional interference that does not take into account the prior art approach. As a result, the interference-directional inhibition approach stops the adaptation process when there is no significant directional interference to be eliminated, whereas the prior art approach does not. For example, when there is a weak source signal (eg during voice intermission) and there is almost no directional interference except some uncorrelated noise (such as noise due to wind or mechanical vibrations in the sensor structure), the SNR-based approach would allow the adaptation filter to continue the adaptation due to the small SNR. The process of continuous adaptation is not desirable because there is very little directional interference that must be eliminated first., and the adaptation process seeks in vain new filter weights to eliminate uncorrelated noise, which frequently results in the cancellation of the component of the source signal of the received signal. In contrast, the inhibition mechanism based on directional interference will inhibit the adaptation process in such a case because the resistance of the directional interference as referenced in the average energy of the reference channel will become smaller than the average of the reference channel. Normalized main channel energy, producing a positive normalized energy difference. The adaptation process is inhibited as a result until some directional interference is eliminated. Figure 9 shows the selective frequency restriction adaptation filter together with the difference unit in accordance with the preferred embodiment of the present invention. The frequency selective restriction matching filter 101 includes a finite impulse response (FIR) filter 102, an LMS weight update unit 103 and a frequency selective weight restriction unit 104. In an alternative mode, the infinite impulse response filter (IIR) can be used instead of the FIR filter. A flat frequency reference channel 105 passes through the FIR filter 102 whose filter weights are adjusted to produce a cancellation signal 106 that closely approximates the component of the current interference signal that is present in the main channel 107. In a preferred embodiment, the main channel is obtained from the main channel matrix unit after a delay in order to synchronize the main channel with the cancellation signal. In general, there is a delay between the main channel and the cancellation signal because the cancellation signal is obtained by the processing reference channels through the extra delay stages, ie the fading filters and the adaptation filters. In an alternative mode, the main channel directly from the matrix unit of the main channel can be used if the delay is not significant. A difference unit 108 subtracts the cancellation signal 106 from the main channel 107 to generate an output signal 109. The adaptation filter 101 adjusts the weights of the filter W] _- Wn, to minimize the energy of the output signal. When the filter weights are set, the output signal 109 generates the source signal essentially free of the component of the current interference signal because the cancellation signal 106 closely follows the track of the component of the interference signal. The output signal 109 is sent to the D / A output unit to produce an analog output signal. The output signal 109 is also used to adjust the weights of the adaptation filter in order to further reduce the component of the interference signal. There are many techniques in order to continuously update the values of the filter weights. The preferred modality uses the least squares average (LMS) algorithm which minimizes the value of the quadratic mean of the difference between the main channel and the cancellation signal but in an alternative mode, other algorithms such as Minimum Recursive Square can also be used. (RLS). Under the LMS algorithm, the adaptation filter weights are updated according to the following: Wp (n + 1) = Wp (n) + 2 μ r (np) e (n) where n is a time index discreet; Wp is the filter weight p-th of the adaptation filter; e (n) is a signal of difference between the signal of the main channel and the cancellation signal; r (n) is a reference channel; and μ is an adaptation constant that controls the speed of adaptation. Figure 10 illustrates a preferred embodiment of the frequency selective weight restriction unit. The selective weight control unit 110 in frequency includes a Fast Fourier Transform unit 112, a set of frequency reservoirs 114, a set of truncation units 115, a set of storage cells 116 and a unit 117 of Fast Fourier Transformation. Reverse (IFFT), connected in series. The FFT unit 112 receives the weights 111 of the adaptation filter and carries out the FFT of the weights of the filter 111 to obtain values 113 for frequency representation. The frequency representation values are then divided into a set of frequency bands and stored in the frequency repositories 114a to 114h. Each frequency repository stores the frequency representation values within a specific bandwidth assigned to each deposit. The values represent the operation of the adaptation filter with respect to a specific frequency component of the source signal. Each of the truncation units 115a to 115h compares the frequency representation values with the threshold assigned to each deposit, and truncates the values if they exceed the threshold. The truncated frequency representation values are temporarily stored at 116a-116h before the IFFT unit 117 converts them again to the new filter weight values 118. In addition to the inhibition mechanism based on directional interference, the frequency selective weight restriction unit also controls the adaptation process based on the frequency spectrum of the received source signal. Once the adaptation filter begins to work, the change of operation at the filter output, better or worse, becomes drastic. Uncontrolled adaptation can quickly lead to drastic degradation of performance. The weight restriction mechanism is based on the observation that a large increase in the weight values of the adaptation filter suggests signal leakage. If the adaptation filter works properly, there is no need for the filter to increase the filter weights at large volumes. But if the filter is not working properly, filter weights tend to grow to large values. One way to curve the growth is to use a simple truncating mechanism to truncate the values of the filter weights to predetermined threshold values. In this way, even when the energy of the total signal can be high enough to activate the inhibition mechanism, the weight restriction mechanism can still prevent signal leakage. For narrow band signals, such as a voice signal or a tonal signal, which has its spectral density of energy concentrated in a narrow frequency range, the signal leakage may not manifest itself in a large growth of the filter weight values in the time domain. However, filter weight values in the frequency domain will indicate some increase because they represent the operation of the adaptation filter in response to a specific presence component of the source signal. In the frequency selective weight restriction unit it detects that condition by detecting a large increase in the frequency representation values of the filter weights. By truncating the frequency representation values in the narrow frequency band of interest and transforming them back in the time domain again, the unit acts to prevent signal leakage involving narrow band signals. The system described herein can be implemented using commercially available digital signal processing (DSP) systems such as the analog device of the 2100 series.

Figure 11 shows a flow chart illustrating the operation of a program for a DSP processor in accordance with a preferred embodiment of the present invention. After the program starts in step 100, the program starts the registers and the indicators as well as the stabilizers (step 110). The program then waits for an interruption of a sampling unit that requests the processing of the samples received from the sensor array (step 120). When the sampling unit sends an interrupt (step 131) that the samples are ready, the program reads the sample values (step 130) and stores the values (step 140). The program filters the stored values using a routine that implements a derived delay line and stores the filtered input values (step 141). The program then retrieves the filtered input values (151) and the coefficients of the main channel matrix (step 152) to generate a main channel (step 150) by multiplying the two and to store the result (step 160). The program retrieves the filtered input values (step 171) and the coefficients of the reference channel matrix (step 172) to generate a reference channel (reference channel # 1) by multiplying the two (step 170) and to store the result (step 180). Steps 170 and 180 are repeated to generate all other reference channels (step 190). The program retrieves one of the reference channels (step 201) and the decolorization filter coefficients for the corresponding reference channel (step 202) in order to generate a flat frequency reference channel by multiplying the two (step 200) and store the result (step 210). Steps 200 and 210 are repeated for all other reference channels (step 220). The program retrieves one of the flat frequency reference channels (step 231) and the coefficients of the adaptation filter (step 232) to generate the cancellation signal (step 230) by multiplying the two and to store the result (step 240). Steps 230 and 240 are repeated for all other reference channels in order to generate more cancellation signals (step 250). The program recovers cancellation signals (step 262-263) to subtract them from the main channel (retrieved in step 261) to cancel the component of the reference signal in the main channel (step 260). The output is sent to a D / A unit to produce the signal without interference in analog form (step 264). The output is also stored (step 270). The program calculates the energy or power of a sample in the reference channel (step 281) and retrieves an average of the energy from the old reference channel (step 282). The program multiplies the energy of the sample by a and the old energy average by (1-a) and adds them (step 280) and stores the result as a new energy average (step 290). This process is repeated for all other reference channels (step 300) and the total sum of the energy averages of all the reference channels is stored (step 310). The program multiplies the energy of a sample of the main channel (recovered in step 321) by means of an average of the old energy of the main channel (recovered in step 322) by (1-a), sum the same (step 320) and stores them as an average energy number of the main channel (step 330). The program then multiplies the energy of the main channel with a threshold to obtain a normalized average power of the main channel (step 340). The program subtracts the total energy average of the reference channel (retrieved in step 341) from the average normalized energy of the main channel to produce a difference (step 350). If the difference is positive, the program goes back to step 120 where it simply waits for the other samples. If the difference is negative, the program supports a Weight update routine. The program calculates a new filter weight by adding [2 times the adaptation constant x of the reference channel sample (retrieved in step 361) by the output (retrieved in step 362)] to the old filter weight (retrieved in step 363) to update the weight (step 370) and store the result (step 370). The program performs the FFT of the new filter weights to obtain its frequency representation (step 380). The frequency representation values are divided into several frequency bands and are stored in a frequency deposit set (step 390). The frequency representation values in each deposit are compared to a threshold associated with each frequency deposit (step 400). If the values exceed the threshold, the values are truncated to the threshold (step 410). The program performs the IFFT to convert the truncated frequency representation values back into filter weight values (step 420) and stores them (step 430). The program repeats the weight update routine, steps 360 to 430, for all other reference channels and associated adaptation filters (step 440). The program then returns to step 120 to wait for an interruption for a new round of processing samples (step 450). Although the invention has been described with reference to the preferred embodiments, it is not limited to those embodiments. It will be appreciated for those skilled in the art that modifications can be made to the structure and form of the invention without departing from its spirit and scope which is stopped in the following claims.

Claims (44)

- R E I V I N D-I C A C I O N E S:

1. An adaptation system for processing digital input data representing signals containing a source signal from an axis signal source relative to a sensor array, as well as interference signals from sources of interference placed outside the axis of the signal. signal source and to produce digital output data representing the source signal with reduced interference signals relative to the source signal, comprising: a main channel matrix unit for generating a main channel of the digital input data , the main channel represents signals received in the direction of the signal source and having a component of the signal source and a component of the interference signal; a reference channel matrix unit for generating at least one reference channel from the digital input data, each reference channel represents signals received in directions other than those of the signal source; at least one matching filter having weights of the matching filter, connected to receive signals from the reference channel matrix unit in order to generate a cancellation signal that approaches components of the interference signals of the main channel; a difference unit connected to receive the signals from the main channel matrix unit, and at least one adaptation filter for generating the digital output data with the cancellation signal of the main channel; at least one matching filter also connected to receive the digital output costs and including a weight updating means to find new filter weight values of at least one matching filter such that the difference between the main channel and cancellation signal is minimized; and a weight restriction means for truncating the new filter weight values to predetermined threshold values, when each of the new filter weight values exceeds the corresponding threshold value.

2. The system of claim 1, further comprising at least one fading filter for filtering at least one reference channel so as to have a frequency spectrum whose magnitude is essentially flat over a predetermined frequency range.

3. The system of claim 1, further comprising an inhibiting means connected to receive the signals from the matrix unit of the main channel and the reference channel matrix unit, to calculate the energy of the main channel and the energy of minus one reference channel, and to generate an inhibition signal towards the weight update means when a normalized energy difference between the main channel and at least one reference channel is positive.

4. The system of claim 1, wherein the sensors are microphones.

5. An adaptation system for processing digital input data representing signals containing a source signal from a signal source on the axis in relation to a sensor array as well as interference signals from sources of interference placed outside of the axis from the signal source, and to produce digital output data representing the source signal with reduced interference signals relative to the source signal, comprising: a main channel matrix unit for generating a main channel from the digital input data, the main channel represents the signals received in the direction of the signal source and having a source signal component and an interference signal component; a reference channel matrix unit for generating at least one reference channel from the digital input data, each reference channel represents the signals received in directions other than those of the signal source; at least one adaptation filter having adaptation filter weights connected to receive the signals from the matrix unit of the reference channel in order to generate a cancellation signal that approximates the component of the interference signal of the main channel; a difference unit connected to receive the signals from the matrix unit of the main channel and at least one adaptation filter, to generate the digital output data by subtracting the cancellation signal from the main channel; at least one adaptation filter which is also connected to receive the data from the digital output and which includes a weight updating means to find the new filter weight values, of at least one adaptation filter in such a manner, That - minimizes the difference between the main channel and the cancellation signal; and a weight restriction means for converting the new values of the filter weight into frequency representation values by truncating the frequency representation values to predetermined threshold values and converting them again to weights of the adaptation filter.

6. The system of claim 5, further comprising at least one discoloration filter for filtering at least one reference channel so as to have a frequency spectrum whose magnitude is essentially planar through a predetermined frequency range.

The system of claim 5, further comprising an inhibiting means connected to receive the signals from the main channel matrix unit and the reference channel matrix unit, to calculate the energy of the main channel and the energy of the main channel. at least one reference channel and to generate an inhibition signal to the weight updating means when the normalized energy difference between the main channel and at least one reference channel is positive.

8. The system of claim 5, wherein the sensors are microphones.

9. An adaptive system for receiving a source signal from the signal source on the axis relative to the system as well as interference signals from the interference sources placed off the axis of the signal source and to produce an output signal with reduced interference signals in relation to the source signal, comprising: a formation of spatially distributed sensor sensors, each to receive the interference source signals; a sampling unit connected to receive the signals from the formation of sensors in order to convert these signals in digital form; a main channel matrix unit connected to receive the sampling unit signals, in order to generate signals representing the main channel received in the direction of the signal source, the main channel having a source signal component and a component of the interference signal; a reference channel matrix unit connected to receive the signals of the sampling unit, "in order to generate at least one reference channel, each reference channel represents the received signals in directions other than that of the source of signals, at least one adaptation filter having adaptation filter weights connected to receive signals from the reference channel matrix unit, in order to generate a cancellation signal that approximates the component of the channel interference signal main, a difference unit connected to receive the signals from the main channel matrix unit and at least one adaptation filter, to subtract the cancellation signal from the main channel in order to create a digital output signal; from digital to analog output to convert the digital output signal in analog form, at least one adaptation filter that is also connected or to receive the digital output signal from the resistance unit and including a weight updating means to find the new filter weight values of at least one adaptation filter in such a way as to minimize the difference between the main channel and cancellation signal; and a weight restriction means for truncating the new filter weight values to threshold values - predetermined when each filter weight value exceeds the corresponding threshold value.

The system of claim 9, further comprising at least one discoloration filter for filtering at least one reference channel so as to have a frequency spectrum whose magnitude is essentially planar over a predetermined frequency range.

The system of claim 9, further comprising an inhibiting means connected to receive the signals of the matrix unit of the main channel and the matrix unit of the reference channel, to calculate the energy of the main channel and the energy of at least one reference channel, and to generate an inhibition signal to the weight updating means when the normalized energy difference between the main channel and at least one reference channel.

The system of claim 9, further comprising a delay means for delaying the main channel so that the main channel is synchronized with the cancellation signal before the difference unit subtracts the cancellation signal from the main channel.

13. The system of claim 9, wherein the sensors are microphones.

The system of claim 13, wherein the microphones are omnidirectional microphones.

15. The system of claim 13, wherein the microphones are unidirectional microphones.

The system of claim 9, wherein the matrix unit of the main channel includes beam-forming means for spatially filtering the signals from the sampling unit to exhibit the highest sensitivity to the signal source.

The system of claim 9, wherein the matrix unit of the reference channel includes beam-forming means for spatially filtering the signals from the sampling unit in order to exhibit the lowest sensitivity to the signal source.

The system of claim 9, wherein at least one adaptive filter comprises a finite impulse response filter for generating the cancellation signal.

The system of claim 9, wherein at least one adaptation filter comprises an infinite impulse response filter for generating the cancellation signal.

20. The system of claim 9, wherein the weight updating means uses the minimum quadratic mean algorithm where the value of the quadratic mean of the difference between the main channel and the cancellation signal is minimized.

21. The adaptation system for receiving a source signal from a signal source between the axis relative to the system as well as the interference signals from the off-axis interference sources from the signal source _ and to produce an output signal with reduced interference signals in relation to the source signal, comprising: a formation of spatially distributed sensor sensors, each to receive these interference source signals; a sampling half connected to receive the signals from the formation of sensors, in order to convert these signals into digital form; a main channel matrix unit, connected to receive the signals from the sampling unit, to generate signals representative of the main channel in the direction of the signal source, the main channel having a source signal component and a signal component of interference; a reference channel matrix unit connected to receive the signals from the sampling unit, in order to generate at least one reference channel, each reference channel represents the signals received in directions other than those of the signal source; at least one adaptation filter having adaptive filter-weights connected to receive the signals of the reference channel matrix unit in order to generate a cancellation signal that approximates the component of the interference signal of the main channel; a difference unit connected to reduce the signals from the matrix unit of the main channel and at least one adaptation filter to subtract the cancellation signal from the main channel in order to generate a digital output signal; a digital to analog output converter to convert the digital output signal in analog form; at least one adaptive filter is also connected to receive the digital output signal from the difference unit and includes a weight updating means to find the new filter weight values of at least one adaptation filter in such a way that the difference between the main channel and the cancellation signal is minimized; and a weight restriction means for restricting the operation of the adaptation filter by converting the new weight values of the filter into frequency representation values, truncating the frequency representation values to predetermined threshold values, and converting them back to filter weight of adaptation.

The system of claim 21, wherein the weight restriction means comprises: a Fast Fourier Transform unit for generating frequency representation values of the new filter weight values; a frequency deposit set, each frequency repository for storing the frequency representation values for a frequency band assigned to each frequency deposit; a set of means of truncating each connected to the corresponding frequency deposit to truncate the frequency representation values stored in each frequency deposit up to a predetermined threshold value, if the frequency representation values exceed the threshold value associated with each frequency deposit; and a reverse Fast Fourier Transformation unit connected to the set of truncating means, to convert the values of the set of truncating means back to the weights of the adaptation filter.

The system of claim 21, further comprising at least one decolorization filter for filtering at least one reference channel so as to have a frequency spectrum whose magnitude is essentially planar over a predetermined frequency range.

The system of claim 21, further comprising an inhibiting means connected to receive the signals from the main channel matrix unit and the reference channel matrix unit, to calculate the energy of the main channel and the energy of the main channel. at least one reference channel and to generate an inhibition signal to the weight updating means, when the normalized energy difference between the main channel and at least one reference channel is positive.

25. The system of claim 21, wherein the sensors are microphones.

26. The system of claim 21, wherein the main channel matrix unit includes beam-forming means for spatially filtering the signals from the sampling unit in order to exhibit the highest sensitivity to the signal source.

The system of claim 21, wherein the matrix unit of the reference channel includes beam-forming means for spatially filtering the signals from the sampling unit in order to exhibit the lowest sensitivity to the signal source.

The system of claim 21, wherein the weight updating means uses the minimum quadratic mean algorithm in which the value of the quadratic mean of the difference between the main channel and the cancellation signal is minimized.

29. A method for processing digital input data representing signals that contain a source signal from a signal source on the axis from a sensor array as well as interference signals from interference sources placed off-axis from the source. signal source, and to produce digital output data representing the source signal with reduced interference signals relative to the source signal, comprising the steps of: generating a main channel of the digital input data, the main channel represents the signals received in the direction of the signal source and has a source signal component and an interference signal component; generating at least one reference channel from the digital input data, each reference channel representing the signals received in directions other than those of the signal source; filtering at least one reference channel using the weight values of the filter to generate a cancellation signal that approaches the component of the interference signal in the main channel; generate the digital output data by subtracting the cancellation signal from the main channel; derive new filter weight values so that the difference between the main channel and the cancellation signal is minimized; and truncating the new filter weight values to predetermined threshold values when each of the new filter weight values exceeds. the corresponding threshold value.

30. The method of claim 29, further comprising the step of filtering at least one reference channel such that it has an essentially flat frequency spectrum.

31. The method of claim 29, further comprising the step of inhibiting the generation of the cancellation signal when the normalized energy difference between the main channel and at least one reference channel is positive.

32. A method for processing the digital input data representing signals containing a source signal from a signal source on the axis from a sensor array as well as the interference signal from the interference sources placed off the axis from the signal source and to produce digital output data representing the source signal with reduced interference signals relative to the source signal, comprising the steps of: generating a main channel of the digital input data, the signals representing the main channel received in the direction of the signal source and having a source signal component and an interference signal component; generating at least one reference channel from the digital input data, each reference channel represents signals received in directions other than those of the signal source; filtering at least one reference channel using the weight values of the filter to generate a cancellation signal that approximates the component of the interference signal in the main channel; 1 - . 1 - generate the digital output data by subtracting the cancellation signal from the main channel; deriving the new values of the weight of the filter so that the difference between the main channel and the cancellation signal is minimized; and restricting the new weight values of the filter by converting the new weight values of the filter into frequency representation values, truncating the frequency representation values to predetermined threshold values and converting them back to the weight values of the filter.

The method of claim 32, wherein the restriction of the new filter weight values comprises: generating frequency representation values of the new filter weight values; dividing the frequency representation values into a plurality of frequency deposits; truncating the frequency representation values in each frequency deposit that exceed a predetermined threshold value associated with each frequency deposit; and convert the frequency representation values back to the filter weight values.

34. The method of claim 33, wherein the generation of the frequency representation is carried out using the Fast Fourier Transfomation and the conversion again is carried out using the Inverse Fast Fourier Transformation.

35. The method of claim 32, further comprising the step of filtering at least one reference channel such that it has an essentially flat frequency spectrum.

36. The method of claim 32, further comprising the step of inhibiting the generation of the cancellation signal when a normalized energy difference between the main channel and at least one reference channel is positive.

37. A method for receiving a source signal from a signal source as well as interference signals and sources of interference and for producing an output signal with reduced interference signals relative to the source signal, comprising the steps of: receiving analog signals from a formation of spatially distributed sensor sensors; take samples of the analog signals to convert them in digital form; generating a channel representing the signals received in the direction of the signal source, the main channel having a component of the source signal and a component of the interference signal; generating at least one reference channel, each reference channel represents signals received in directions other than those of the signal source; filter at least one reference channel using filter weight values to generate a cancellation signal that approximates the component of the interference signal in the main channel; generate a digital output signal by subtracting the cancellation signal from the main channel; convert the digital output signal into analog form; derive the new filter weight values in such a way as to minimize the difference between the main channel and the cancellation signal; and truncating the new filter weight values to predetermined threshold values when each of the new filter weight heaps exceeds the corresponding threshold value.

38. The method of claim 37, further comprising the step of filtering at least one reference channel such that it has an essentially flat frequency spectrum.

39. The method of claim 37, further comprising the step of inhibiting the generation of the cancellation signal when a difference of the normalized energy between the main channel and at least one reference channel is positive.

40. A method for receiving a source signal from a signal source as well as interference signals from the interference sources and for producing an output signal with reduced interference signals relative to the source signal, comprising the steps of : receiving analog signals from a sensor array of spatially distributed sensors; take samples of the analog signals to convert them in digital form; generating a main channel representing the signals received in the direction of the signal source, the main channel having a source signal component and a component of the interference signal; generating at least one reference channel, each reference channel represents the signals received in directions other than those of the signal source; filtering at least one reference channel using filter weight values to generate a cancellation signal that approximates the component of the interference signal in the main channel; generate a digital output signal by subtracting the cancellation signal from the main channel; convert the digital output signal in analog form; derive the new filter weight values in such a way as to minimize the difference between the main channel and the cancellation signal; and restricting the new filter weight values from converting them back into filter weight values into frequency representation values, truncating the frequency representation values to predetermined threshold values and converting them back into filter weight values.

41. The method of claim 40, wherein the restriction of the new weight values of the filter comprises: generating frequency representation values of the new filter weight values; dividing the frequency representation values into a plurality of frequency deposits; truncate the frequency representation values in each frequency deposit if they exceed a predetermined threshold value associated with each frequency deposit; and converting the frequency representation value back into filter weight values.

42. The method of claim 41, wherein the generation of the frequency representation values is carried out using the Fourier Transform. Fast and convert them back to filter weight values which is carried out using the Reverse Fast Fourier Transformation.

43. The method of claim 40, further comprising the step of filtering at least one reference channel such that it has an essentially flat frequency spectrum.

44. The method of claim 40, further comprising the step of inhibiting the generation of the cancellation signal when the difference of the normalized energy between the main channel and at least one reference channel is positive.