JP2000165301A - Identification method for unknown system by band split adaptive filter, recoding medium recording identification program and identification device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an identification method of an unknown system by a coefficient update algorithm for a band split adaptive filter with a short convergence time by using a limited arithmetic amount or a memory capacity. SOLUTION: The identification device consists of an unknown system 108 that receives an input signal, a band split filter 102 that splits the received signal into a plurality of sub band reference input signals, an independent adaptive filter 106 that receives each sub band reference input signal, a coefficient update section 105 that updates its coefficient, a 2nd band split filter 109 that divides an output signal of the unknown system into a plurality of sub band main input signals, and an error signal calculation section 111 that calculates a sub band error signal being a difference between a plurality of sub band main input signals and a plurality of output signals of the adaptive filter, and any of a plurality of coefficient update algorithms is selected adaptively once for a period based on the reference input signal, the main input signal or an error signal and the coefficient of the adaptive filter is updated thereby.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for identifying an unknown system such as a transmission path and a spatial acoustic coupling path, and a computer-readable recording medium and apparatus recording an identification program.

[0002]

2. Description of the Related Art An echo canceller, a noise canceller, a howling canceller, an adaptive equalizer, and the like are known as applications of identification of an unknown system using an adaptive filter. Here, a conventional technique will be described by taking an acoustic echo canceller that removes an acoustic echo leaking from a speaker to a microphone in a spatial acoustic coupling path as an example.

An echo canceller generates a pseudo echo corresponding to a transmission signal by using an adaptive filter having a larger number of taps than an impulse response length of an echo path, so that sound leaking from a speaker to a microphone in a spatial acoustic coupling path. Operate to suppress echo.
At this time, each filter coefficient of the adaptive filter is corrected using an error signal and a far-end signal obtained by subtracting a pseudo echo from a mixed signal in which an echo and a near-end signal are mixed. As typical examples of such an adaptive filter coefficient updating algorithm, an “NLMS algorithm” and a “FRLS algorithm” are known.

[0004] The impulse response length of an acoustic space in which an acoustic echo canceller is actually used depends on the physical dimensions of the acoustic space and the reflectivity of walls and the like. For example, assuming a room used for a video conference or the like, the impulse response length is 1
It is not uncommon for the number of taps to be as large as 000 taps and sometimes even thousands, which is extremely difficult to realize from the viewpoint of the amount of calculation and the hardware scale. In order to solve such a problem of mainly an increase in the amount of calculation, a band division type (sub-band type) adaptive filter has been proposed.

Next, the conventional band-division acoustic echo canceller will be described with reference to FIG. FIG. 7 is a block diagram showing a configuration of a conventional band-division acoustic echo canceller. In FIG. 7, the NLMS algorithm is used as the coefficient updating algorithm. First, the sound input signal sent from the far end 701 is
Divides into a plurality of bands (sub-bands) to generate a sub-band input signal. Each sub-band input signal is supplied to a decimation unit 7
03i (i = 0, 1,..., K−1), which are thinned out to 1 / Li and independent NLMS algorithms 704i
(I = 0, 1,..., K−1) and the adaptive filter 70
5i (i = 0, 1,..., K−1). Usually, Li ≦ K is set.

On the other hand, the output of the unknown system 706 to be identified, that is, the acoustic echo in the case of the acoustic echo canceller, is also divided into a plurality of bands by another band division filter 707 having exactly the same characteristics as the band division filter 702. After being subjected to the sub-band acoustic echo, the thinning unit 708i
(I = 0, 1,..., K−1) is thinned out to 1 / Li. The error signal calculation unit 709i generates a subband error signal that is the difference between the subband pseudo echo output from the adaptive filter and the thinned subband acoustic echo.
The NLMS algorithm 704i updates the coefficient using the subband error signal.

The sub-band error signal is supplied to an interpolation unit 710
After being interpolated by Li (i = 0, 1,..., K−1) to be Li times, supplied to the band synthesis filter 711, band-synthesized, and transmitted to the output unit 712. Therefore, the signal obtained at the output unit 712 is such that if each sub-band error signal is sufficiently small, that is, if the sub-band acoustic echo is sufficiently suppressed in each sub-band, the acoustic echo is suppressed even when viewed as a whole band. Signal.

[0008]

However, in the above-described method for identifying an unknown system using the conventional band division adaptive filter, the same algorithm for updating the coefficients of the sub-band adaptive filter is used in all bands. When it is necessary to identify an unknown system with a limited amount of computation or memory, it is difficult to realize real-time operation using a high-performance coefficient update algorithm with a large amount of computation and memory. was there.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-mentioned conventional problems. An unknown system using a coefficient updating algorithm of a band division adaptive filter having a short convergence time in a limited amount of computation or memory. It is an object of the present invention to provide a recording medium and an apparatus in which an identification method and an identification program are recorded.

[0010]

According to the present invention, in order to solve the above-mentioned problems, in identifying an unknown system using a band division adaptive filter, information obtained from any of a reference input signal, a main input signal, and an error signal is used. Is used to select a coefficient update algorithm of the adaptive filter of each subband once in a certain period, and uses a coefficient update algorithm having a high convergence speed within a limited amount of computation or memory. That is, the present invention provides at least two or more types of coefficient updating algorithms for updating the coefficients of the adaptive filter,
By selecting the coefficient updating algorithm adaptively once in a certain period, the unknown system can be efficiently identified by the band division type adaptive filter that can efficiently identify the unknown system within the limited amount of calculation and memory. An identification method is obtained.

In order to solve the above-mentioned problems, the present invention provides at least two or more types of coefficient updating algorithms for updating coefficients of an adaptive filter, and obtains information obtained from any of a reference input signal, a main input signal, and an error signal. Once, once in a certain period, adaptively select the coefficient update algorithm,
The adaptive filter coefficients are updated by the selected coefficient update algorithm, and the unknown system is identified by the band division adaptive filter that can efficiently identify the unknown system with a limited amount of computation and memory. A recording medium on which the program has been recorded is obtained.

According to the present invention, in order to solve the above-mentioned problems, the coefficient updating unit prepares at least two or more types of coefficient updating algorithms for updating the coefficients of the adaptive filter, and selects one of the reference input signal, the main input signal, and the error signal. Using the obtained information, once in a certain period, the coefficient updating algorithm is adaptively selected, and the coefficient of the adaptive filter is updated by the selected coefficient updating algorithm. Also, an unknown system identification device using a band division adaptive filter capable of efficiently identifying an unknown system in the amount of memory can be obtained.

[0013]

The method for identifying an unknown system according to the first aspect of the present invention supplies an input signal to the unknown system and simultaneously divides the band to generate a plurality of sub-band reference input signals. The band reference input signal is supplied to an independent adaptive filter for each subband and a coefficient updating unit for updating the coefficient of the adaptive filter, and the output signal of the unknown system is divided into bands to generate a plurality of subband main input signals. Calculating a difference between each of the plurality of sub-band main input signals and an output signal of the adaptive filter to output a sub-band error signal; and updating the coefficient of the adaptive filter using the calculated sub-band error signal. An unknown system identification method comprising the steps of: preparing at least two or more types of coefficient update algorithms for updating the coefficients of the adaptive filter;
Once in a certain period, the coefficient updating algorithm is adaptively selected. At least two or more types of coefficient updating algorithms are used for updating the coefficient of the sub-band adaptive filter, and the algorithm is used once in a certain period. By adaptively selecting a coefficient updating algorithm, an unknown system can be effectively identified with a limited amount of calculation and memory.

According to a second aspect of the present invention, in the method for identifying an unknown system, the step of adaptively selecting a coefficient updating algorithm is performed based on information obtained by analyzing the subband reference input signal. Selecting a coefficient updating algorithm of the coefficient updating unit of each band, based on information obtained by analyzing each subband reference input signal. As a result, an unknown system can be effectively identified within a limited amount of computation and memory.

According to a third aspect of the present invention, in the method for identifying an unknown system, the step of adaptively selecting the coefficient updating algorithm is performed based on information obtained by analyzing the subband main input signal. A coefficient updating algorithm of a coefficient updating unit of a band is selected, and a coefficient updating algorithm of a coefficient updating unit of each subband is selected based on information obtained by analyzing a plurality of subband main input signals. This has the effect that an unknown system can be effectively identified within a limited amount of computation and memory.

According to a fourth aspect of the present invention, in the method for identifying an unknown system, the step of adaptively selecting the coefficient updating algorithm is a difference between the subband main input signal and the output signal of the adaptive filter. Based on the information obtained by analyzing the sub-band error signal, the coefficient updating algorithm of the coefficient updating unit of each sub-band is selected, and the information obtained by analyzing each sub-band error signal is By selecting the coefficient update algorithm of the coefficient update unit of each subband based on the above, there is an effect that an unknown system can be effectively identified with a limited amount of calculation and memory.

According to a fifth aspect of the present invention, a recording medium storing an unknown system identification program supplies an input signal to an unknown system, and simultaneously divides a band to generate a plurality of subband reference input signals. Each sub-band reference input signal is supplied to an independent adaptive filter for each sub-band and to a coefficient updating unit for updating the coefficient of the adaptive filter, and the output signal of the unknown system is divided into a plurality of sub-band main input signals. , Calculating a difference between each of the plurality of sub-band main input signals and an output signal of the adaptive filter to output a sub-band error signal, and performing at least two or more types of coefficient update algorithms for updating the coefficients of the adaptive filter. Prepare,
Using the calculated sub-band error signal, once in a certain period, adaptively selecting the coefficient update algorithm, and each step of updating the coefficient of the adaptive filter by the selected coefficient update algorithm. Yes, at least two or more types of coefficient updating algorithms are used for updating the coefficients of the sub-band adaptive filter, and the coefficient updating algorithm is adaptively selected once in a certain period, so that a limited amount of calculation and a limited amount of memory are required. It has an effect that an unknown system can be effectively identified in the system.

According to a sixth aspect of the present invention, there is provided a recording medium recording an unknown system identification program, wherein the step of adaptively selecting a coefficient updating algorithm includes information obtained by analyzing the sub-band reference input signal. , The coefficient updating algorithm of the coefficient updating unit of each subband is selected, and the coefficient updating of the coefficient updating unit of each subband is performed based on the information obtained by analyzing the subband reference input signal. By selecting an algorithm, there is an effect that an unknown system can be effectively identified within a limited amount of calculation and memory.

According to a seventh aspect of the present invention, there is provided a recording medium on which an unknown system identification program is recorded, wherein the step of adaptively selecting a coefficient updating algorithm includes information obtained by analyzing the sub-band main input signal. , The coefficient updating algorithm of the coefficient updating unit of each subband is selected, and the coefficient of the coefficient updating unit of each subband is determined based on information obtained by analyzing a plurality of subband main input signals. Selecting an update algorithm has the effect that an unknown system can be effectively identified within a limited amount of computation and memory.

According to another aspect of the present invention, there is provided a recording medium on which an unknown system identification program is recorded, wherein the step of adaptively selecting a coefficient updating algorithm includes the sub-band main input signal and the output signal of the adaptive filter. Based on the information obtained by analyzing each subband error signal that is the difference between
The coefficient updating algorithm of the coefficient updating unit of each subband is selected, and the coefficient updating algorithm of the coefficient updating unit of each subband is selected based on information obtained by analyzing each subband error signal. This has the effect that an unknown system can be effectively identified within a limited amount of computation and memory.

According to a ninth aspect of the present invention, there is provided an apparatus for identifying an unknown system, wherein the unknown system to be identified is divided into a plurality of subbands to generate a plurality of subband reference input signals. A filter, an independent adaptive filter for each of the subbands receiving the respective subband reference input signals, and a coefficient updating unit for updating the coefficients of the adaptive filter, and dividing the output signal of the unknown system into a plurality of subbands A second band division filter that generates a plurality of sub-band main input signals, and a difference between the two signals by supplying the plurality of sub-band main input signals and a plurality of output signals of the adaptive filter. An error signal calculating unit for calculating a sub-band error signal, wherein the coefficient of the adaptive filter is calculated using the calculated sub-band error signal in each of the coefficient updating units. In the identification device for an unknown system that is to be updated, the coefficient updating unit prepares at least two or more types of coefficient updating algorithms for updating the coefficients of the adaptive filter, and adaptively executes the coefficient updating algorithm once in a certain period. By selecting at least two or more types of coefficient updating algorithms for updating the coefficients of the adaptive filter of each subband, and selecting the coefficient updating algorithm adaptively once in a certain period, This has the effect that an unknown system can be effectively identified within a limited amount of calculation and memory.

According to a tenth aspect of the present invention, there is provided an apparatus for identifying an unknown system, comprising: a plurality of reference input signal analyzers for analyzing information of a plurality of subband reference input signals; And a coefficient updating algorithm selecting section for selecting a coefficient updating algorithm of the coefficient updating section for each sub-band based on the analysis result of the signal analyzing section, based on the analysis result of each sub-band reference input signal. By selecting the coefficient updating algorithm of the coefficient updating unit of each subband, there is an effect that an unknown system can be effectively identified with a limited amount of calculation and memory.

An apparatus for identifying an unknown system according to claim 11 of the present invention comprises: a plurality of main input signal analysis units for analyzing information of a plurality of subband main input signals;
And a coefficient updating algorithm selecting section for selecting a coefficient updating algorithm of the coefficient updating section for each subband based on the analysis result of the main input signal analyzing section. By selecting the coefficient update algorithm of the coefficient update unit of each subband based on the result, it is possible to effectively identify an unknown system with a limited amount of computation and memory.

According to a twelfth aspect of the present invention, there is provided an apparatus for identifying an unknown system, comprising: an error signal analyzer for analyzing information of a plurality of subband error signals; and an analyzer for analyzing the subband error signal once in a certain period. A coefficient updating algorithm selecting unit for selecting a coefficient updating algorithm of a coefficient updating unit for each subband based on the result, and a coefficient updating unit for each subband based on an analysis result of each subband error signal. By selecting the coefficient updating algorithm, it is possible to effectively identify an unknown system with a limited amount of computation and memory.

According to a method for identifying an unknown system according to a thirteenth aspect of the present invention, the power of a signal in each sub-band is used as information for adaptively selecting the coefficient updating algorithm. The use of power as information when selecting a coefficient updating algorithm has an effect that an unknown system can be effectively identified within a limited amount of calculation and memory with a simple calculation.

According to a method for identifying an unknown system according to a fourteenth aspect of the present invention, a root mean square value of a signal of each subband in a certain period is used as information for adaptively selecting the coefficient updating algorithm. By using the root mean square value of a signal during a certain period as information when selecting a coefficient update algorithm, a simple operation can be performed to achieve an effective unknown system with a limited operation amount and memory amount. It has the effect of being able to be identified.

According to a fifteenth aspect of the present invention, there is provided a recording medium recording an unknown system identification program such that the power of a signal in each subband is used as information for adaptively selecting the coefficient updating algorithm. The use of power as information when selecting a coefficient update algorithm has the effect that simple calculations can be used to effectively identify unknown systems within a limited amount of computation and memory. .

A recording medium on which the program for identifying an unknown system according to claim 16 of the present invention is recorded may include, as information for adaptively selecting the coefficient updating algorithm, a root-mean-square signal of each subband in a certain period. The value is used, and by using the root mean square value of the signal for a certain period as information when selecting the coefficient update algorithm, the calculation can be performed with a simple calculation and the limited calculation amount and the memory amount are effective. It has an effect that a systematic unknown system can be identified.

An unknown system identification apparatus according to claim 17 of the present invention uses signal power in each subband as information when adaptively selecting the coefficient updating algorithm, The use of power as information when selecting a coefficient updating algorithm has an effect that an unknown system can be effectively identified within a limited amount of calculation and memory with a simple calculation.

The apparatus for identifying an unknown system according to claim 18 of the present invention uses the root mean square value of the signal of each subband in a certain period as information when adaptively selecting the coefficient updating algorithm. By using the root mean square value of a signal during a certain period as information when selecting a coefficient update algorithm, a simple operation can be performed to achieve an effective unknown system with a limited operation amount and memory amount. It has the effect of being able to be identified.

[0031] In the method for identifying an unknown system according to claim 19 of the present invention, the coefficient updating algorithm of the coefficient updating unit includes an NLMS algorithm (learning identification method) and a FRLS.
The NLMS algorithm (learning identification method) and the FRLS algorithm are used as the coefficient updating algorithm of the coefficient updating unit. Since it can be assigned to a band that is considered to be large in amount, it has an effect that a high convergence speed can be obtained as a whole.

According to a twentieth aspect of the present invention, there is provided a recording medium storing an unknown system identification program, wherein an NLMS algorithm (learning identification method) and a FRLS algorithm are used for a coefficient updating algorithm of the coefficient updating unit. NLMS algorithm (learning identification method) and FRL are used as the coefficient updating algorithm of the coefficient updating unit.
By using the S algorithm, the amount of calculation is large,
Since the fast convergence rate FRLS algorithm can be assigned to the band that seems to have the largest amount of echo,
This has the effect of obtaining a high convergence speed as a whole.

According to a twenty-first aspect of the present invention, in the apparatus for identifying an unknown system, an NLMS algorithm (a learning identification method) and a FRLS
By using the NLMS algorithm (learning identification method) and the FRLS algorithm as the coefficient updating algorithm of the coefficient updating unit, the FRLS algorithm that requires a large amount of computation but has the fastest convergence speed is the most echo Since it can be assigned to a band that is considered to be large in amount, it has an effect that a high convergence speed can be obtained as a whole.

According to a method of identifying an unknown system according to a twenty-second aspect of the present invention, the selection of the coefficient updating algorithm of the coefficient updating unit in each subband is performed once every integer number of taps per subband. In the case where the FRLS algorithm, which is generally said to converge within 3 × (the number of taps per band), is used as the coefficient updating algorithm, the subroutine using the FRLS algorithm is used. It is possible to select a coefficient updating algorithm after the band has converged, and it is possible to stably and surely identify an unknown system.

According to a twenty-third aspect of the present invention, in the recording medium storing the unknown system identification program, the selection of the coefficient updating algorithm of the coefficient updating unit in each subband is performed by an integer multiple of the number of taps per subband. And the number of adaptations is generally three.
When the FRLS algorithm which is said to converge within x (the number of taps per band) is used for the coefficient updating algorithm, the coefficient updating algorithm can be selected after the subband using the FRLS algorithm converges,
This has the effect that the unknown system can be identified stably and reliably.

In the identification system for unknown systems according to a twenty-fourth aspect of the present invention, the selection of the coefficient updating algorithm of the coefficient updating unit in each subband is performed once every integer multiple of the number of taps per subband. In the case where the FRLS algorithm, which is generally said to converge within 3 × (the number of taps per band), is used as the coefficient updating algorithm, the subroutine using the FRLS algorithm is used. It is possible to select a coefficient updating algorithm after the band has converged, and it is possible to stably and surely identify an unknown system.

Hereinafter, based on the attached drawings, FIGS. 1 to 6,
An embodiment of the present invention will be described in detail. Here, as an example of the unknown system, an acoustic echo canceller that removes an acoustic echo leaking from a speaker to a microphone in a spatial acoustic coupling path will be described. Specifically, the number of band divisions is thinned out into eight divisions and six, and the coefficient updating algorithm of the adaptive filter is NLM because of the restrictions on the amount of computation and the amount of memory.
It is assumed that the S algorithm has six bands and the FRLS algorithm with a short convergence time has two bands.

(Embodiment 1) Next, referring to FIG.
An apparatus and method for identifying an unknown system using a band division adaptive filter according to Embodiment 1 of the present invention will be described. FIG. 1 is a block diagram showing a configuration of an unknown system identification device using a band division adaptive filter according to Embodiment 1 of the present invention. In the first embodiment, an acoustic echo canceller will be described as an example of an unknown system identification device.

In FIG. 1, 101 is a far end, 102 is a band division filter, 103i is a thinning unit, 104i is a reference input signal analysis unit, 105i is a coefficient update unit, 106i is an adaptive filter, 107 is a coefficient update algorithm selection unit, 1
08 is an unknown system, 109 is a band division filter as a second band division filter, 110i is a thinning unit, 11
1i is an error signal calculation unit, 112i is an interpolation unit, 113 is a band synthesis filter, 114 is an output unit, 115 is a speaker,
116 is a microphone.

Next, the operation of the apparatus for identifying an unknown system using the band division adaptive filter according to the first embodiment will be described with reference to FIG. First, an input signal sent from the far end 101 is output to the unknown system 108 through the speaker 115 and is also input to the band division filter 102. The band division filter 102 divides the input signal into eight bands by using a method such as a polyphase filter, and generates a sub-band reference input signal. Each of the divided subband reference input signals is supplied to a decimation unit 103i (i =
0, 1,..., 7), which are thinned out to 1/6, and independent reference input signal analysis units 104i (i = 0, 1,...)
., 7) and the coefficient updating unit 105i (i =
0, 1,..., 7) and the adaptive filter 106i (i =
0, 1,..., 7). The reference input signal analysis unit 104i analyzes the power, the sum of the absolute values, the average of the absolute values, and the mean square value using one or more samples of the reference input signal.

In the coefficient updating algorithm selecting section 107,
The result analyzed by the reference input signal analysis unit 104i is compared, and a coefficient update algorithm in the coefficient update unit 105i for each subband is selected. Specifically, when the analysis method of the reference input signal analysis unit 104i is power calculation, the FRLS algorithm is assigned to the top two bands in descending order of power, and the NLMS algorithm is assigned to the other bands. The result of the coefficient update algorithm selection unit 107 is obtained when the number of adaptations exceeds 3 × (the number of taps per band), which is said to be necessary for the FRLS algorithm to converge, or in the previous analysis result. Fast adaptive filter coefficient updating algorithm, F in this case
This is reflected after the subbands to which the RLS algorithm is assigned have sufficiently converged. The FRLS algorithm requires a large amount of calculation but has a fast convergence speed. Therefore, if the FRLS algorithm is assigned to a band in which the amount of echo is considered to be the largest, the convergence speed is increased as a whole.

On the other hand, the unknown system 10 to be identified
8, the acoustic echo in the case of the acoustic echo canceller is also input to the microphone 116, and then divided into a plurality of bands by the band division filter 109 different from the band division filter 102, and the sub-band which is the main input signal After becoming an acoustic echo, the thinning unit 110i (i = 0, 1,.
.., 7), it is thinned out to 1/6.

The error signal calculator 111i (i = 0, 1,...)
.., 7) calculate a sub-band error signal that is a difference between the sub-band pseudo echo output from the adaptive filter 106i and the thinned-out sub-band acoustic echo. The coefficient updating unit 105i updates the coefficients of the adaptive filter 106i by using the subband error signal and the coefficient updating algorithm selected by the coefficient updating algorithm selecting unit 107. This sub-band error signal is supplied to the interpolation unit 112
After being interpolated 6 times by i (i = 0, 1,..., 7) and supplied to the band synthesizing filter 113 and band synthesized by a method such as a polyphase filter, the output unit 11
4 is transmitted. Therefore, the signal obtained at the output unit 114 is such that if each sub-band error signal is sufficiently small, that is, if the sub-band acoustic echo is sufficiently suppressed in each sub-band, the acoustic echo is suppressed even when viewed as a whole band. Signal.

The operation of the apparatus for identifying an unknown system by the band division adaptive filter according to the first embodiment has been described above with reference to the configuration diagram of FIG. 1. Next, with reference to FIG. A description will be given of the flow of processing when the method of identifying an unknown system is realized by a software program. FIG. 2 is a flowchart showing a method of identifying an unknown system in the acoustic echo canceller shown in FIG. In FIG.
Step 210 is a process for the reference input signal, Step 220 is a process for the main input signal, Step 250 is a process for the error signal, and Steps 260, 270,
And 280 are processes of the coefficient update algorithm selection unit 107.

First, at step 211, the far end 101
, The input data is divided into sub-bands by the band division filter 102, and a sub-band reference input signal is output (step 212).
In 3i, each sub-band reference input signal is thinned out to 1/6 (step 213), and the thinned-out sub-band reference input signal is analyzed by the reference input signal analysis unit 104i into power or the like (step 213).

Next, the microphone 1 is connected to the speaker 115.
The acoustic echo inputted to the sub-band 16 (step 221) is divided into sub-bands by the band division filter 109 to output a sub-band main input signal (step 222). 6 is thinned out (step 223). An error signal calculator 111i calculates a subband error signal from the decimated subband main input signal and the output of the adaptive filter of each subband (step 230), and the coefficient updating unit 105i converts the subband error signal into Then, the coefficients of the adaptive filter 106i are updated by the coefficient updating algorithm selected by the coefficient updating algorithm selecting unit 107 (step 240).

On the other hand, the output sub-band error signal is interpolated by a factor of 6 in the interpolation section 112i (step 251), supplied to the band synthesis filter 113, synthesized (step 252), and output from the output section 114. (Step 253). Next, in step 260, the FRLS
It is determined whether or not the number of adaptations is greater than 3 × (the number of taps per band) which is said to be necessary for the algorithm to converge. If no, the process returns to step 211; At 270, a coefficient update algorithm is selected using the analysis result of the reference input signal analysis unit 104i. Next, the process proceeds to step 280 to reset the number of adaptations, and returns to step 211.

(Embodiment 2) Next, referring to FIG.
An apparatus and method for identifying an unknown system using a band division adaptive filter according to Embodiment 2 of the present invention will be described. FIG. 3 is a block diagram showing a configuration of an unknown system identification device using a band division adaptive filter according to Embodiment 2 of the present invention. In the second embodiment, an acoustic echo canceller will be described as an example of an unknown system identification device.

In FIG. 3, reference numeral 301 denotes a far end, 302 denotes a band division filter, 303i denotes a thinning unit, 304i denotes a main input signal analyzing unit, 305i denotes a coefficient updating unit, 306i denotes an adaptive filter, 307 denotes a coefficient updating algorithm selecting unit, 30
8 is an unknown system, 309 is a band division filter as a second band division filter, 310i is a thinning unit, 311
i is an error signal calculation unit, 312i is an interpolation unit, 313 is a band synthesis filter, 314 is an output unit, 315 is a speaker,
Reference numeral 16 denotes a microphone.

Next, with reference to FIG. 3, the operation of the apparatus for identifying an unknown system using the band division adaptive filter according to the second embodiment will be described. First, the input signal sent from the far end 301 is output to the unknown system 308 through the speaker 315, and at the same time, the band division filter 302
Is input to The band division filter 302 divides the input signal into eight bands using a method such as a polyphase filter, and generates a sub-band reference input signal. Each subband reference input signal is supplied to a decimation unit 303i (i = 0, 1,.
.., 7), the coefficients are reduced to 1/6 and independent coefficient updating units 305i (i = 0, 1,..., 7) and adaptive filters 306i (i = 0, 1,. ).

On the other hand, the unknown system 30 to be identified
8, the acoustic echo in the case of an acoustic echo canceller is also input to the microphone 316, and then divided into a plurality of bands by a band division filter 309 different from the band division filter 302, and After becoming an acoustic echo, the thinning section 310i (i = 0, 1,.
.., 7), it is thinned out to 1/6.

The decimated sub-band acoustic echoes are converted into independent main input signal analyzers 304i (i = 0, 1,.
, 7) are supplied to the error signal calculation unit 311i, which is independent of each other. The main input signal analysis unit 304i uses one or more samples of the main input signal to calculate power, sum of absolute values,
Analysis is performed on feature values such as an average value of absolute values and a root mean square value. The coefficient updating algorithm selecting section 307 compares the results analyzed by the main input signal analyzing section 304i and selects an algorithm of the coefficient updating section 305i for each subband. Specifically, when the analysis method of the main input signal analysis unit 304i is power calculation, the FRLS algorithm is applied to the top two bands in descending order of power, and the NLM is applied to the other bands.
Assign the S algorithm.

The result of the coefficient update algorithm selecting section 307 is obtained by the 3 × (the number of taps per band) which is said to be necessary for the number of adaptations to be converged by the FRLS algorithm, or the result of the previous analysis. This is reflected, for example, after the subbands to which the high-speed adaptive filter coefficient updating algorithm, in this case the FRLS algorithm is assigned, have sufficiently converged. Error signal calculator 311i (i = 0, 1, 1)
.., 7) calculate a subband error signal which is a difference between the subband pseudo echo output from the adaptive filter 306i and the thinned subband acoustic echo.

Using the subband error signal, coefficient updating section 305i updates the coefficients of adaptive filter 306i according to the coefficient updating algorithm selected by the coefficient updating algorithm selecting section. This sub-band error signal is interpolated six times by an interpolation unit 312i (i = 0, 1,..., 7) and then supplied to a band synthesis filter 313.
After being subjected to band synthesis by a method such as a polyphase filter, it is transmitted to the output unit 314. Therefore, the output unit 31
The signal obtained in 4 is a signal in which the echo is suppressed even in the entire band if the sub-band error signal is sufficiently small, that is, if the sub-band echo is sufficiently suppressed in each sub-band.

The operation of the apparatus for identifying an unknown system using the band division adaptive filter according to the second embodiment has been described above with reference to the configuration diagram of FIG. 3. Next, with reference to FIG. A description will be given of the flow of processing when the method of identifying an unknown system is realized by a software program. FIG. 4 is a flowchart showing a method for identifying an unknown system in the acoustic echo canceller shown in FIG. In FIG.
Step 410 is processing for a reference input signal, step 420 is processing for a main input signal, step 450 is processing for an error signal, and steps 460, 470,
And 480 are processes of the coefficient update algorithm selection unit 307.

First, in step 411, the far end 301
, The input data is divided into subbands by the band division filter 302, and a subband reference input signal is output (step 412).
At 3i, each subband reference input signal is thinned out to 1/6 (step 413), and output to the coefficient update unit 305i and the adaptive filter 306i.

Next, the speaker 315 is connected to the microphone 3
The acoustic echo input to the sub-band 16 (step 421) is divided into sub-bands by the band division filter 309 to output a sub-band main input signal (step 422). 6 is thinned out (step 423). The decimated sub-band main input signal is analyzed into power or the like by the main input signal analyzer 304i (step 424). An error signal calculator 311i calculates a sub-band error signal from the decimated main input signal of each sub-band and the output signal of the adaptive filter of each sub-band (step 430). , The coefficient of the adaptive filter 306i is updated by the coefficient updating algorithm selected by the coefficient updating algorithm selecting unit 307 (step 440).

On the other hand, the output sub-band error signal is interpolated 6 times (step 451) by interpolation unit 312i, supplied to band synthesis filter 313, synthesized (step 452), and output from output unit 314. (Step 453). Next, in step 460, it is determined whether or not the number of adaptations is greater than 3 × (the number of taps per band) which is said to be necessary for the FRLS algorithm to converge. If it is high, in step 470, a coefficient update algorithm is selected using the analysis result of the main signal analysis unit 304i. Next, the process proceeds to step 480 to reset the number of adaptations, and returns to step 411.

(Embodiment 3) Next, referring to FIG.
An apparatus and method for identifying an unknown system using a band division adaptive filter according to Embodiment 3 of the present invention will be described. FIG. 5 is a block diagram showing a configuration of an unknown system identification device using a band division adaptive filter according to Embodiment 3 of the present invention. In the third embodiment, an acoustic echo canceller will be described as an example of an unknown system identification device.

In FIG. 5, reference numeral 501 denotes a far end, 502 denotes a band division filter, 503i denotes a thinning-out unit, 504i denotes an error signal analyzing unit, 505i denotes a coefficient updating unit, 506i denotes an adaptive filter, 507 denotes a coefficient updating algorithm selecting unit, and 508.
Is an unknown system, 509 is a band division filter as a second band division filter, 510i is a thinning unit, 511i
Is an error signal calculation unit, 512i is an interpolation unit, 513 is a band synthesis filter, 514 is an output unit, 515 is a speaker, 51
Reference numeral 6 denotes a microphone.

Next, with reference to FIG. 5, the operation of the unknown system identification apparatus using the band division adaptive filter according to the third embodiment will be described. First, an input signal sent from the far end 501 is output to the unknown system 508 through the speaker 515 and at the same time to the band division filter 502. The band division filter 502 divides the input signal into eight bands by a method such as a polyphase filter, and generates a sub-band reference input signal. Each subband reference input signal is output to a decimation unit 503i (i = 0, 1,...).
, 7), which are thinned out to １ ／, and independent coefficient updating units 505i (i = 0, 1,..., 7) and adaptive filters 506i (i = 0, 1,..., 7). Supplied to

On the other hand, the unknown system 50 to be identified
8, the acoustic echo in the case of the acoustic echo canceller is also input to the microphone 516, and is then divided into a plurality of bands by a band division filter 509 different from the band division filter 502, and is a sub-band which is a main input signal. After an acoustic echo, the thinning unit 510i (i = 0, 1,.
.., 7), it is thinned out to 1/6.

The error signal calculator 511i (i = 0, 1,...)
.., 7) calculate a subband error signal which is a difference between the subband pseudo echo output from the adaptive filter 506i and the decimated subband acoustic echo. The sub-band error signals are output from independent error signal analyzers 504i.
(I = 0, 1,..., 7) and supplied to independent coefficient updating units 505i. The error signal analysis unit 504i analyzes the power, the sum of the absolute values, the average value of the absolute values, the mean square value, and the like using one or more samples of the error signal. The coefficient update algorithm selection section 507 compares the results analyzed by the error signal analysis section 504i and selects an algorithm of the coefficient update section 505i for each subband.

Specifically, when the analysis method of the error signal analyzer 504i is power calculation, the FRLS algorithm is assigned to the upper two bands in descending order of power, and the NLMS algorithm is assigned to the other bands. The result of the coefficient update algorithm selection unit 507 indicates that the number of adaptations is FRLS.
3 × (the number of taps per band) which is said to be necessary for the algorithm to converge, or a high-speed adaptive filter coefficient update algorithm based on the previous analysis result. In this case, FRLS This is reflected after the subbands to which the algorithm is assigned have sufficiently converged. The coefficient update unit 505i uses this subband error signal to calculate
The coefficient of the adaptive filter 506i is updated by the coefficient updating algorithm selected by the coefficient updating algorithm selecting unit 507.

The sub-band error signal is supplied to an interpolation unit 512
After being interpolated six times by i (i = 0, 1,..., 7) and supplied to the band synthesizing filter 513 and band-synthesized by a method such as a polyphase filter, the output unit 51
4 is transmitted. Therefore, the signal obtained at the output unit 514 is such that if each sub-band error signal is sufficiently small, that is, if the sub-band acoustic echo is sufficiently suppressed in each sub-band, the acoustic echo is suppressed even when viewed as a whole band. Signal.

The operation of the apparatus for identifying an unknown system by the band division adaptive filter according to the third embodiment has been described above with reference to the configuration diagram of FIG. 5. Next, with reference to FIG. A description will be given of the flow of processing when the method of identifying an unknown system is realized by a software program. FIG. 6 is a flowchart showing a method for identifying an unknown system in the acoustic echo canceller shown in FIG. In FIG.
Step 610 is processing for a reference input signal, step 620 is processing for a main input signal, step 650 is processing for an error signal, and steps 660, 670,
And 680 are processes of the coefficient update algorithm selection unit 507.

First, at step 611, the far end 501
, The input data is divided into sub-bands by the band division filter 502, and a sub-band reference input signal is output (step 612).
In 3i, each subband reference input signal is thinned out to 1/6 (step 613), and output to the coefficient update unit 505i and the adaptive filter 506i.

Next, the speaker 515 to the microphone 5
The acoustic echo inputted to the sub-band 16 (step 621) is divided into sub-bands by the band division filter 509 to output a sub-band main input signal (step 622). 6 (step 623). An error signal calculating unit 511i calculates a subband error signal from the decimated main input signal of each subband and the output of the adaptive filter of each subband (step 630), and the coefficient updating unit 505i uses the subband error signal. And coefficient updating algorithm selecting unit 5
The coefficient of the adaptive filter 506i is updated by the coefficient updating algorithm selected in step 07 (step 64).
0).

The sub-band error signal calculated by the error signal calculating section 511i is analyzed into power or the like by the error signal analyzing section 504i (step 651), and the result is sent to the coefficient updating algorithm selecting section 507. On the other hand, the sub-band error signal output from the error signal analyzer 504i is interpolated six times by the interpolator 512i (step 65).
2) It is supplied to the band synthesizing filter 513, synthesized (step 653), and output from the output unit 514 (step 654). Next, at step 660, it is determined whether or not the number of adaptations is greater than 3 × (the number of taps per band) which is said to be necessary for the FRLS algorithm to converge. If it is high, in step 670, a coefficient update algorithm is selected using the analysis result of the error signal analyzer 504i. Next, the process proceeds to step 680, where the number of adaptations is reset, and the process returns to step 611.

The method of identifying an unknown system using the above-described band division type adaptive filter can be realized by a software program in addition to hardware, and the software program is not limited to FD, CD-ROM, and any recording such as IC. Even in a medium, it can be recorded and executed.

[0071]

According to the present invention, at least two or more types of coefficient updating algorithms are prepared for updating the coefficients of the adaptive filter, and the reference input signal, the main input signal,
Using information obtained from signals such as error signals, adaptively selecting a coefficient updating algorithm of the adaptive filter of each subband once in a certain period, and thereby updating the coefficients of the adaptive filter. An unknown system identification method using a band-division adaptive filter that can efficiently identify an unknown system among the calculated amount of computation and memory is obtained.

According to the present invention, in particular, at least two or more kinds of coefficient updating algorithms are prepared for updating coefficients of an adaptive filter, and information obtained from signals such as a reference input signal, a main input signal, and an error signal is used for a certain period. Once, the coefficient updating algorithm is adaptively selected, and the coefficients of the adaptive filter are updated by the selected coefficient updating algorithm. A recording medium that records an unknown system identification program using a band division type adaptive filter capable of identifying an unknown system is obtained.

According to the present invention, in particular, the coefficient updating unit prepares at least two or more types of coefficient updating algorithms for updating the coefficients of the adaptive filter, and uses information obtained from any of the reference input signal, the main input signal, and the error signal. And once in a certain period,
The coefficient updating algorithm is adaptively selected, and the coefficients of the adaptive filter are updated by the selected coefficient updating algorithm. Efficient identification of an unknown system in a limited operation amount and memory amount is achieved. An apparatus for identifying an unknown system by a possible band division type adaptive filter is obtained.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of an unknown system identification device using a band division adaptive filter according to Embodiment 1 of the present invention;

FIG. 2 is a flowchart showing a method of identifying an unknown system in the acoustic echo canceller shown in FIG. 1,

FIG. 3 is a block diagram showing a configuration of an unknown system identification device using a band division adaptive filter according to Embodiment 2 of the present invention;

FIG. 4 is a flowchart showing a method of identifying an unknown system in the acoustic echo canceller shown in FIG. 3,

FIG. 5 is a block diagram showing a configuration of an unknown system identification device using a band division adaptive filter according to Embodiment 3 of the present invention;

6 is a flowchart showing a method of identifying an unknown system in the acoustic echo canceller shown in FIG. 5,

FIG. 7 is a block diagram showing a configuration of an unknown system identification device using a conventional band division adaptive filter.

[Explanation of symbols]

101 far end, 102 band division filter, 103i
Decimation unit, 104i Reference input signal analysis unit, 105i
Coefficient updating unit, 106i adaptive filter, 107 coefficient updating algorithm selecting unit, 108 unknown system, 109
Band division filter, 110i Thinning section, 111i
Error signal calculation unit, 112i interpolation unit, 113 band synthesis filter, 114 output unit, 115 speaker, 116
Microphone 210 Processing for reference input signal, 211 Reference data input, 212 Band division, 213 decimation, 214
Analysis, 220 processing on main input signal, 221 reference data input, 222 band division, 223 decimation, 23
0 error calculation, 240 coefficient update, 250 processing for error signal, 251 interpolation, 252 band synthesis, 253
Error output, 260 Determines whether or not the number of adaptations exceeds 3 × (the number of taps of the adaptive filter per band), 2
70 Selection of coefficient update algorithm using analysis result of reference input 280 Reset of number of adaptations 301
Far end, 302 band division filter, 303i decimation unit, 304i main input signal analysis unit, 305i coefficient update unit, 306i adaptive filter, 307 coefficient update algorithm selection unit, 308 unknown system, 309 band division filter, 310i decimation unit 311i error signal Calculation unit, 312i interpolation unit, 313
i band synthesis filter, 314 output unit, 315 speaker, 316 microphone 410 processing for reference input signal, 411 reference data input, 412 band division, 413 decimation, 420
Processing for main input signal, 421 Main data input, 42
2 band division, 423 decimation, 424 analysis, 430
Error calculation, 440 coefficient update, 450 processing for error signal, 451 interpolation, 452 band synthesis, 453
Error output, 460 Judgment as to whether or not the number of adaptations exceeds 3 × (the number of taps of the adaptive filter per band); 47
0 Selection of a coefficient update algorithm using the result of analyzing the main input, 480 Reset of the number of adaptations, 501 Far end, 502 Band division filter, 503i thinning unit,
504i error signal analyzer, 505i coefficient updater, 5
06i adaptive filter, 507 coefficient update algorithm selection unit, 508 unknown system, 509 band division filter, 510i thinning unit, 511i error signal calculation unit, 512i interpolation unit, 513 band synthesis filter,
14 output unit, 515 speaker, 516 microphone, 610 processing for reference input signal, 611 reference data input, 612 band division, 613 thinning, 62
0 processing for main input signal, 621 main data input,
622 band division, 623 decimation, 630 error calculation, 640 coefficient update, 650 processing for error signal, 652 interpolation, 653 band synthesis, 654 error output, 651 analysis, 660 The number of adaptations is 3 × (tap of adaptive filter per band) Number), 670 Selection of a coefficient update algorithm using the result of analyzing the error signal, 680 Reset of the number of adaptations 701 Far end, 702 Band division filter, 703i
Decimation unit, 704i coefficient updating unit, 705i adaptive filter, 706 unknown system, 707 band division filter, 708i decimation unit, 709i error signal calculation unit,
710i interpolation unit, 711 band synthesis filter, 712
Output unit, 713 speaker, 714 microphone

Claims (24)

[Claims] An adaptive filter for supplying a plurality of sub-band reference input signals to an unknown system by supplying the input signal to an unknown system and dividing the sub-band reference input signals into independent sub-bands for each of the sub-bands; The coefficient is supplied to a coefficient updating unit that updates the coefficient of the unknown system, to generate a plurality of sub-band main input signals by dividing the output signal of the unknown system,
Calculating a difference between the plurality of sub-band main input signals and an output signal of the adaptive filter to output a sub-band error signal;
An unknown system identification method comprising a step of updating coefficients of the adaptive filter using the calculated sub-band error signal, wherein at least two or more types of coefficient updating algorithms are prepared for updating the coefficients of the adaptive filter. A method for identifying an unknown system using a band division adaptive filter, wherein the coefficient update algorithm is adaptively selected once during a period. 2. The method according to claim 1, wherein the step of adaptively selecting a coefficient updating algorithm selects a coefficient updating algorithm of a coefficient updating unit of each subband based on information obtained by analyzing the subband reference input signal. 2. A method for identifying an unknown system by using a band division adaptive filter according to claim 1. 3. The step of adaptively selecting a coefficient updating algorithm includes selecting a coefficient updating algorithm of a coefficient updating unit of each subband based on information obtained by analyzing the subband main input signal. 2. A method for identifying an unknown system by using a band division adaptive filter according to claim 1. 4. The step of adaptively selecting a coefficient updating algorithm is based on information obtained by analyzing each subband error signal which is a difference between the subband main input signal and the output signal of the adaptive filter. 2. The method for identifying an unknown system by a band division adaptive filter according to claim 1, wherein a coefficient updating algorithm of a coefficient updating unit of each subband is selected. 5. The method of claim 1, wherein the input signal is supplied to an unknown system, and is simultaneously band-divided to generate a plurality of sub-band reference input signals.
The sub-band reference input signal is supplied to an independent adaptive filter for each sub-band and a coefficient updating unit for updating the coefficient of the adaptive filter, and the output signal of the unknown system is divided into a plurality of sub-band main inputs. Generating a signal, calculating a difference between each of the plurality of sub-band main input signals and an output signal of the adaptive filter and outputting a sub-band error signal, and updating at least two or more types of coefficient updating algorithms for the adaptive filter. Using the calculated sub-band error signal, adaptively selecting the coefficient update algorithm once in a certain period, and updating the coefficients of the adaptive filter by the selected coefficient update algorithm. Computer-readable recording medium for recording unknown system identification program by band division type adaptive filter characterized by . 6. The step of adaptively selecting a coefficient update algorithm includes selecting a coefficient update algorithm of a coefficient update unit of each subband based on information obtained by analyzing the subband reference input signal. A computer-readable recording medium recording an unknown system identification program by the band division adaptive filter according to claim 5. 7. The step of adaptively selecting a coefficient updating algorithm includes selecting a coefficient updating algorithm of a coefficient updating unit of each subband based on information obtained by analyzing the subband main input signal. A computer-readable recording medium recording an unknown system identification program by the band division adaptive filter according to claim 5. 8. The step of adaptively selecting a coefficient updating algorithm based on information obtained by analyzing each subband error signal which is a difference between the subband main input signal and an output signal of the adaptive filter. 6. A computer-readable recording medium recording an unknown system identification program by a band division adaptive filter according to claim 5, wherein a coefficient updating algorithm of a coefficient updating unit of each subband is selected. 9. An unknown system to be identified, a band division filter that divides an input signal into a plurality of subbands to generate a plurality of subband reference input signals, and receives each of the subband reference input signals. An independent adaptive filter for each subband and a coefficient updating unit for updating coefficients of the adaptive filter; and a second unit for dividing the output signal of the unknown system into a plurality of subbands and generating a plurality of subband main input signals. And an error signal calculating unit that supplies the plurality of sub-band main input signals and the plurality of output signals of the adaptive filter and calculates a difference between two signals to calculate a sub-band error signal. In the identification device of the unknown system, so as to update the coefficient of the adaptive filter using the calculated sub-band error signal in each of the coefficient update unit, The coefficient updating unit prepares at least two or more kinds of coefficient updating algorithms for updating the coefficients of the adaptive filter, and adaptively selects the coefficient updating algorithm once during a certain period. Identification system for unknown systems using adaptive filters. 10. A plurality of reference input signal analyzers for analyzing information of a plurality of subband reference input signals, and said coefficient of each subband based on an analysis result of said reference input signal analyzer once during a certain period. 10. The apparatus according to claim 9, further comprising a coefficient updating algorithm selecting unit that selects a coefficient updating algorithm of the updating unit. 11. A plurality of main input signal analyzers for analyzing information of a plurality of sub-band main input signals, each of which comprises:
10. The band division type adaptive filter according to claim 9, further comprising: a coefficient updating algorithm selecting unit that selects a coefficient updating algorithm of the coefficient updating unit for each subband based on an analysis result of the main input signal analyzing unit. For identifying unknown systems. 12. An error signal analysis unit for analyzing information of a plurality of sub-band error signals, and a coefficient updating algorithm of a coefficient updating unit for each sub-band based on an analysis result of the sub-band error signal once in a certain period. 10. An apparatus for identifying an unknown system using a band division adaptive filter according to claim 9, further comprising: a coefficient update algorithm selection unit for selecting a coefficient. 13. The unknown by the band division adaptive filter according to claim 2, wherein the power of a signal in each subband is used as information when adaptively selecting the coefficient updating algorithm. How to identify the system. 14. The band division according to claim 2, wherein a root mean square value of a signal of each subband in a certain period is used as information when the coefficient updating algorithm is adaptively selected. Of unknown system using adaptive filter. 15. The unknown by the band division type adaptive filter according to claim 6, wherein the power of a signal in each subband is used as information for adaptively selecting the coefficient updating algorithm. A computer-readable recording medium recording a system identification program. 16. The band division according to claim 6, wherein a root mean square value of a signal of each subband in a certain period is used as information for adaptively selecting the coefficient updating algorithm. A computer-readable recording medium that records an unknown system identification program using a type adaptive filter. 17. The unknown by the band division adaptive filter according to claim 10, wherein the power of a signal in each subband is used as information for adaptively selecting the coefficient updating algorithm. System identification device. 18. The method according to claim 1, wherein a mean square value of a signal of each subband in a certain period is used as information for adaptively selecting the coefficient updating algorithm.
An apparatus for identifying an unknown system using the band division adaptive filter according to 0, 11 or 12. 19. A coefficient updating algorithm of the coefficient updating unit, wherein an NLMS algorithm (learning identification method) and a FRLS
3. The method according to claim 1, wherein an algorithm is used.
13. The method for identifying an unknown system by using the band division adaptive filter according to 3, 4, 13, or 14. 20. An NLMS algorithm (learning identification method) and a FRLS
7. The method according to claim 5, wherein an algorithm is used.
A computer-readable recording medium recording an unknown system identification program by the band division adaptive filter according to 7, 8, 15, or 16. 21. An NLMS algorithm (learning identification method) and a FRLS
9. The method according to claim 9, wherein an algorithm is used.
An apparatus for identifying an unknown system using the band division adaptive filter according to any one of 0, 11, 12, 17 and 18. 22. The band division according to claim 1, wherein the selection of the coefficient update algorithm of the coefficient update unit in each subband is performed once every integer number of taps per subband. Of unknown system using adaptive filter. 23. The band division according to claim 5, wherein the selection of the coefficient update algorithm of the coefficient update unit in each subband is performed once per integer multiple of the number of taps per subband. A computer-readable recording medium that records an unknown system identification program using a type adaptive filter. 24. The band division according to claim 9, wherein the selection of the coefficient update algorithm of the coefficient update unit in each subband is performed once every integer multiple of the number of taps per subband. Identification system for unknown systems using adaptive filters.