JP2009267486A - Adaptive equalizer, adaptive equalization method, and recording medium - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an adaptive equalizer and an adaptive equalization method that can stably determine convergence degree even when signal is momentarily changing in power and spectrum like audio or even when noise exists, and to provide a recording medium. <P>SOLUTION: The adaptive equalizer is used for communication equipment, an audio signal processing apparatus etc., and includes a transfer function estimating means of estimating a transfer function to be equalized, a filter means of performing equalization processing, a filter coefficient updating means of updating the filter coefficient of the filter means based upon the estimation result of the transfer function estimating means, a filter coefficient power arithmetic means of calculating the sum of coefficient electric powers having an arbitrary specified number of filer coefficients in a filter coefficient group of the filter means or all the filter coefficients, and a convergence degree determining means of determining the degree of convergence of the filter means based upon the arithmetic result of the filter coefficient power arithmetic means. The update progress state of filter coefficients is directly detected and used to determine the convergence degree. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

The present invention relates to an adaptive equalizer, an adaptive equalization method, and a recording medium for executing the adaptive equalization method used in communication devices, audio signal processing apparatuses, and the like.

  There is an adaptive equalizer (also referred to as “adaptive filter” or “adaptive filter”) as a device used in speech signal processing such as compensation of signal transmission path characteristics in communication equipment and echo and noise removal in speech devices. The adaptive equalizer regards the signal transmission path characteristic, the echo path or the noise path as an unknown system and performs an unknown system estimation operation.

  The basic configuration of the adaptive equalizer will be described below with reference to FIG.

  FIG. 12 is a diagram showing the configuration of a basic adaptive equalizer. In FIG. 12, 1201 is a reference signal, 1202 is an unknown system to be subjected to adaptive estimation, 1203 is an adaptive equalizer that performs an estimation operation of the unknown system 1202, 1204 is a desired signal, 1205 is an output of the adaptive equalizer 1203. A filter output signal 1206 is a subtractor that subtracts the filter output signal 1205 from the desired signal 1204 and outputs a signal that is an error in the estimation operation, and 1207 is an error signal that is an output of the subtractor 1206. The adaptive equalizer 1203 outputs a filter output signal 1205 as a replica candidate signal of the desired signal 1204 that is the output of the unknown system 1202 while performing the estimation operation of the unknown system 1202. Furthermore, the adaptive equalizer 1203 feeds back the state of the error signal 1207 that is the difference between the desired signal 1204 and the filter output signal 1205, thereby performing an adaptive operation to increase the estimation accuracy of the unknown system.

  Next, an application example of the adaptive equalizer will be described with reference to FIG.

  FIG. 13A is a diagram showing a configuration example when the adaptive equalizer is applied as an ANC (active noise control) device for duct noise. Below, a first microphone 1303 installed near the noise generation source, an adaptive equalizer 1203 for estimating noise and generating a filter output signal, and a phase inverter for generating an antiphase signal of the filter output signal 1304 shows an example of an active noise control device 1307 including a speaker 1305 for emitting an estimated anti-phase signal of the noise into a duct, and a second microphone 1306 for collecting the effect of noise reduction. ing. The basic operation of the active noise control device 1307 will be described below.

  Reference numeral 1301 denotes a duct serving as a passage for air conditioning or exhaust gas, and 1302 denotes a duct noise path serving as a transmission path for noise generated in the duct 1301. The adaptive equalizer 1203 in the active noise control device 1307 uses the noise picked up by the first microphone 1303 as a reference signal 1201, and the filter output signal 1205 that estimates the noise estimated to be transmitted through the duct noise path 1302. Is generated. The filter output signal 1205 is converted into an antiphase signal by the subsequent phase inverter 1304 and emitted from the speaker 1305 into the duct to cancel and reduce the duct noise. Further, the second microphone 1306 installed at the subsequent stage of the speaker 1305 collects the noise-reduced signal as an error signal 1207 and feeds it back to the adaptive equalizer 1203, thereby further adapting the operation and further reducing noise. To increase.

  FIG. 13B is a diagram illustrating a configuration example when the adaptive equalizer is applied as a line echo canceller in a communication device. The basic operation of the echo canceller 1309 constituted by the adaptive equalizer 1203 and the subtractor 1206 will be described below.

  When the communication device sends a line transmission signal to the communication line, it may be superimposed on the line reception signal via the line echo path 1308. In this case, an echo canceller 1309 is made up of an adaptive equalizer 1203 that performs an adaptive equalization operation with a line transmission signal as a reference signal 1201 and a line reception signal as a desired signal 1204, and a subtractor 1206 that subtracts the filter output signal 1205 from the desired signal 1204. By applying this, it is possible to output an error signal 1207 as an echo reduction signal. Further, by feeding back an error signal 1207 as an echo reduction signal to the adaptive equalizer 1304, further adaptive operation is advanced, and the echo reduction effect is further enhanced.

  A technique in the case where the adaptive equalizer is used as a line echo canceling device will be described below using (Patent Document 1).

  FIG. 14 is a block diagram showing an adaptive equalizer according to the prior art, and is a device (line echo canceling device) described in (Patent Document 1).

  14, 1401 is an adder, 1402 is an adaptive filter, 1403 is an echo canceller, 1404 is a hybrid circuit, 1405 and 1407 are power calculators, 1406 is a convergence determination unit, 1408 is a delay circuit, 1409 is a reference signal, Reference numeral 1410 denotes an input signal, and 1411 denotes an output signal. Here, the adaptive filter 1402 cancels the line echo component wrapping through the hybrid circuit 1404 in the adder 1401, and the power of the output signal 1411 that is the residual echo component at that time is the first power calculation unit. 1405, the second power calculation unit 1407 calculates the power of the reference signal 1409, and the convergence determination unit 1406 determines the convergence based on the ratio of these powers. The delay circuit 1408 corrects a time lag between the reference signal 1409 and the output signal 1411.

  FIG. 15 is a functional block diagram showing a convergence determination unit 1406 and power calculation units 1405 and 1407 of the conventional adaptive equalizer.

  In FIG. 15, power calculation units 1405 and 1407 have the same configuration, 1501 is a multiplier, 1502 is a coefficient unit, 1503 is an adder, 1504 is a coefficient unit, and 1505 is a delay circuit. The convergence determination unit 1406 includes a divider 1506, a comparison determination control unit 1507, and a threshold setting unit 1508.

The operation of each part configured as above will be described. The output signals of the first and second power calculators 1405 and 1407 indicate the power of the output signal 1411 that is a residual echo component and the power of the reference signal 1409, respectively. Then, the convergence A is calculated by the divider 1406 of the convergence determination unit 1406.
A = (output of first power calculation unit 1405) / (output of second power calculation unit 1407)
The comparison degree control unit 1507 compares the degree of convergence A of the output of the divider 1506 with a predetermined threshold value set in the threshold value setting unit 1508 to determine the degree of convergence.
Japanese Patent Laid-Open No. 08-037480

  In an adaptive equalizer, it is the most basic requirement for control to discriminate and evaluate the degree of adaptation, that is, the degree of convergence. However, in the above conventional adaptive equalizer, the power and spectrum of the adaptive equalizer reference signal 1409 and the output signal 1411 after the equalization are compared with each other so as to determine the degree of convergence. For such signals, the equalization performance of the adaptive equalizer itself greatly depends on the signal power and spectral fluctuations. Therefore, the degree of convergence can be determined stably only by calculating the above power ratio. It had the problem that it was difficult to do. In addition, in the presence of noise, the difference in power is relatively small, so that the degree of convergence cannot be correctly determined.

  With this adaptive equalizer, adaptive equalization method and recording medium, it is possible to stably determine the degree of convergence even for a signal whose power and spectrum fluctuate from moment to moment such as speech. It is required that the degree of convergence can be determined stably even in the presence of noise.

  In order to satisfy this requirement, the present invention can stably determine the degree of convergence even for a signal whose power and spectrum fluctuate from moment to moment, such as voice, and in the presence of noise. An adaptive equalizer that can stably determine the degree of convergence, and a signal that has a power and spectrum that varies from moment to moment, such as speech, can be determined stably. Another object of the present invention is to provide an adaptive equalization method for stably determining the degree of convergence even in the presence of noise, and a recording medium for executing the adaptive equalization method.

  An adaptive equalizer according to the present invention is an adaptive equalizer used in a communication device, an audio signal processing apparatus, and the like, and includes a transfer function estimation unit that estimates a transfer function to be equalized, and a filter that performs equalization processing. Means, a filter coefficient updating means for updating a coefficient value of the filter means based on an estimation result in the transfer function estimating means, and a coefficient power possessed by a specified arbitrary number or all of the filter coefficients in the filter coefficient group of the filter means The filter coefficient power calculating means for calculating the sum, and the convergence degree determining means for determining the degree of convergence of the filter means based on the calculation result of the filter coefficient power calculating means are provided.

  As a result, it is possible to stably determine the degree of convergence even for signals whose power and spectrum fluctuate from moment to moment, such as voice, and to stabilize the degree of convergence even in the presence of noise. Thus, an adaptive equalizer that can be performed is obtained.

  The adaptive equalization method of the present invention is an adaptive equalization method used for communication equipment, an audio signal processing device, etc., and includes a transfer function estimation step for estimating a transfer function to be equalized, and a filter for performing equalization processing. A filter coefficient updating step for updating a coefficient value of the filter step based on an estimation result in the transfer function estimating step, and a coefficient power of a specified arbitrary number or all of the filter coefficients in the filter coefficient group of the filter step A filter coefficient power calculation step for calculating the sum and a convergence determination step for determining the degree of convergence of the filter step based on the calculation result of the filter coefficient power calculation step are provided.

  As a result, it is possible to stably determine the degree of convergence even for a signal whose power and spectrum fluctuate every moment, such as voice, and to perform the determination of the degree of convergence stably even in the presence of noise. An adaptive equalization method is obtained.

  The recording medium of the present invention has a configuration that is a computer-readable recording medium in which a program for executing each step of the adaptive equalization method is recorded.

  Thereby, a recording medium for executing the adaptive equalization method is obtained.

  An adaptive equalizer according to the present invention is an adaptive equalizer used in a communication device, an audio signal processing apparatus, and the like, and includes a transfer function estimation unit that estimates a transfer function to be equalized, and a filter that performs equalization processing. Means, a filter coefficient updating means for updating a coefficient value of the filter means based on an estimation result in the transfer function estimating means, and a coefficient power possessed by a specified arbitrary number or all of the filter coefficients in the filter coefficient group of the filter means The filter coefficient power calculating means for calculating the sum, and the convergence degree determining means for determining the degree of convergence of the filter means based on the calculation result of the filter coefficient power calculating means. Since it is possible to directly detect the update progress of a certain filter coefficient and use it to determine the degree of convergence, it can be used as a voice and so on. -It is possible to stably determine the degree of convergence of the adaptive equalizer even for signals whose spectrum varies, and to stably determine the degree of convergence even in the presence of noise. The advantageous effect is obtained.

  Furthermore, the convergence determination means detects the amount of change in the coefficient power sum calculated by the filter coefficient power calculation means, so that the presence of a signal or noise whose power and spectrum fluctuate momentarily such as voice. Since the update progress of the filter coefficient itself that is not easily affected by fluctuations can be used for the convergence determination, an advantageous effect that the determination of the convergence degree can be performed stably is obtained.

  Furthermore, the convergence degree determination means uses the ratio of the power sum at the previous determination and the power sum at the current determination as the amount of change in the coefficient power sum calculated by the filter coefficient power calculation means, so that There is an advantageous effect that it is possible to flexibly cope with two conflicting demands, that is, a demand for reduction and speeding up judgment and a demand for improvement in judgment accuracy.

  Furthermore, the convergence degree determination means uses the difference between the power sum at the previous determination and the power sum at the current determination as the amount of change in the coefficient power sum calculated by the filter coefficient power calculation means, so that There is an advantageous effect that it is possible to flexibly cope with two conflicting demands, that is, a demand for reduction and speeding up judgment and a demand for improvement in judgment accuracy.

  Furthermore, the convergence determination means calculates the difference between the power sum at the previous determination and the power sum at the current determination, and detects a change in the sign state of the calculation result, thereby detecting a positive sign and a negative sign in the calculation result sign. When the code appears with equal probability, it can be determined that the filter is in a converged state, so it is easier to determine the completion of convergence than simply determining the time change of the power sum and continuing to trace the degree of convergence. The advantageous effect that is possible is obtained.

  Further, the filter coefficient power calculation means reduces the circuit or calculation scale by using the absolute value of each filter coefficient value instead of performing the square calculation of each filter coefficient value in the power calculation of each filter coefficient value. In addition, it is possible to obtain an advantageous effect that it is possible to respond to the demand for speeding up the determination.

  Furthermore, the convergence determination means detects the amount of change in the power sum calculated by the filter coefficient power calculation means in synchronization with the timing at which the filter coefficient update means updates the filter coefficient, so that the filter update operation is appropriately performed. An advantageous effect is obtained that it is possible to perform a stable determination by preventing erroneous determination of the convergence degree assuming that no change in the power sum is recognized at a timing that is not performed.

  The adaptive equalization method of the present invention is an adaptive equalization method used for communication equipment, an audio signal processing device, etc., and includes a transfer function estimation step for estimating a transfer function to be equalized, and a filter for performing equalization processing. Step, a filter coefficient updating step for updating the coefficient value of the filter step based on the estimation result in the transfer function estimating step, and calculating the coefficient power sum of the specified arbitrary number or all the filter coefficients in the filter coefficient group of the filter step A filter coefficient power calculation step to perform, and a convergence degree determination step for determining the degree of convergence of the filter step based on the calculation result of the filter coefficient power calculation step, thereby updating the filter coefficient which is the coefficient update result of the filter coefficient update means Since it is possible to detect the progress directly and use it to determine the degree of convergence, It is possible to stably determine the convergence level of the adaptive equalizer even for signals whose power and spectrum fluctuate from time to time, and to determine the degree of convergence even in the presence of noise. Thus, an advantageous effect can be obtained.

  Further, the convergence determination step detects the amount of change in the coefficient power sum calculated in the filter coefficient power calculation step, so that in the presence of a signal or noise whose power and spectrum fluctuate momentarily such as voice. Since the update progress of the filter coefficient itself that is not easily affected by fluctuations can be used for the convergence determination, an advantageous effect that the determination of the convergence degree can be performed stably is obtained.

  Furthermore, the convergence determination step uses the ratio of the power sum at the previous determination and the power sum at the current determination as the amount of change in the coefficient power sum calculated by the filter coefficient power calculation step. There is an advantageous effect that it is possible to flexibly cope with two conflicting demands, that is, a demand for reduction and speeding up judgment and a demand for improvement in judgment accuracy.

  Furthermore, the convergence determination step uses the difference between the power sum at the previous determination and the power sum at the current determination as the amount of change in the coefficient power sum calculated by the filter coefficient power calculation step, so that There is an advantageous effect that it is possible to flexibly cope with two conflicting demands, that is, a demand for reduction and speeding up judgment and a demand for improvement in judgment accuracy.

  Further, the convergence determination step calculates a difference between the power sum at the previous determination and the power sum at the current determination, and detects a change in the sign state of the calculation result, thereby detecting a positive sign in the sign of the calculation result. When the negative sign appears with equal probability, it can be determined that the filter is in a converged state, so it is easier to determine the completion of convergence than simply determining the power sum over time and tracing the degree of convergence. The advantageous effect that it becomes possible is obtained.

  Further, the filter coefficient power calculation step reduces the circuit or calculation scale by using the absolute value of each filter coefficient value instead of performing the square calculation of each filter coefficient value in the power calculation of each filter coefficient value. In addition, it is possible to obtain an advantageous effect that it is possible to respond to the demand for speeding up the determination.

  Further, the convergence determination step detects the amount of change in the power sum calculated by the filter coefficient power calculation step in synchronization with the timing at which the filter coefficient update step updates the filter coefficient, so that the filter update operation operates appropriately. An advantageous effect is obtained that it is possible to perform a stable determination by preventing erroneous determination of the convergence degree assuming that no change in the power sum is recognized at a timing that is not performed.

  The recording medium of the present invention is a computer-readable recording medium recording a program for executing each step of the adaptive equalization method according to any of the eighth to fourteenth aspects of the invention. An advantageous effect is obtained that the adaptive equalization method according to any of the eighth to fourteenth aspects of the invention can be executed at an arbitrary place at an arbitrary time.

  The present invention stably determines the degree of convergence even for a signal whose power and spectrum fluctuate from moment to moment, such as speech, etc., and stably determines the degree of convergence even in the presence of noise. This object is realized by calculating the coefficient power sum of a specified arbitrary number or all the filter coefficients in the filter coefficient group of the filter means and determining the degree of convergence of the filter means based on the calculation result.

  1st invention made in order to solve the said subject is an adaptive equalizer used for communication equipment, an audio signal processor, etc., and is a transfer function estimating means for estimating a transfer function to be equalized, Filter means for performing equalization processing, filter coefficient update means for updating coefficient values of the filter means based on the estimation result in the transfer function estimation means, and a specified arbitrary number or all of the filter coefficient groups of the filter means The filter coefficient power calculating means for calculating the coefficient power sum of the filter coefficients, and the convergence degree determining means for determining the degree of convergence of the filter means based on the calculation result of the filter coefficient power calculating means. It is possible to directly detect the update progress of the filter coefficient, which is the coefficient update result of the filter coefficient update means, and use it for determining the convergence level. Therefore, it is possible to stably determine the convergence level of the adaptive equalizer even for signals such as speech whose power and spectrum fluctuate from moment to moment, and the degree of convergence even in the presence of noise. It has the effect | action and effect that determination of can be performed stably.

  The second invention made to solve the above-mentioned problem is that the convergence determination means detects the amount of change in the coefficient power sum calculated by the filter coefficient power calculation means, such as voice. Because it is possible to use the update progress of the filter coefficient itself that is not easily affected by fluctuations even in the presence of signals and noise whose power and spectrum fluctuate every moment, it is possible to judge the degree of convergence. It has the effect | action and effect that can be performed stably.

  According to a third aspect of the invention made to solve the above-described problem, the convergence determination unit is configured to calculate the power sum at the previous determination and the power at the current determination as the change amount of the coefficient power sum calculated by the filter coefficient power calculation unit. The ratio of the sum is used, and it is possible to flexibly respond to two conflicting demands: a demand for reducing the circuit or computation scale and speeding up judgment, and a demand for improving judgment accuracy. It has the action and effect of becoming.

  According to a fourth aspect of the invention made to solve the above-described problem, the convergence determination means uses the power sum at the previous determination and the power at the current determination as the change amount of the coefficient power sum calculated by the filter coefficient power calculation means. The difference between the sums is used, and it is possible to flexibly respond to two conflicting demands: a demand for reducing the circuit or computation scale and speeding up judgment, and a demand for improving judgment accuracy. It has the action and effect of becoming.

  According to a fifth aspect of the invention made to solve the above problem, the convergence determination means calculates a difference between the power sum at the previous determination and the power sum at the current determination, and detects a change in the sign state of the calculation result. If the positive and negative signs appear with equal probability in the sign of the operation result, it can be determined that the filter is in a converged state, so simply determine the temporal change in the power sum Thus, it is possible to easily determine the completion of convergence rather than continuously tracing the degree of convergence.

  A sixth invention made to solve the above-mentioned problem is that the filter coefficient power calculation means replaces the square calculation of each filter coefficient value in the power calculation of each filter coefficient value. The absolute value is used, and there is an action and an effect that it becomes possible to cope with a demand for reduction in circuit or calculation scale and speeding up determination.

  According to a seventh aspect of the present invention made to solve the above-mentioned problem, the degree of convergence is determined by the degree of convergence of the power sum calculated by the filter coefficient power calculating means in synchronization with the timing at which the filter coefficient updating means updates the filter coefficient. Stable by preventing the wrong convergence from being judged as no change in power sum is recognized under the timing when the filter update operation is not operating properly It has the effect | action and effect that determination can be performed.

  An eighth invention made to solve the above problem is an adaptive equalization method used for communication equipment, an audio signal processing device, etc., and a transfer function estimation step for estimating a transfer function to be equalized; A filter step for equalization processing, a filter coefficient update step for updating the coefficient value of the filter step based on the estimation result in the transfer function estimation step, and a specified arbitrary number or all of the filter coefficients in the filter coefficient group of the filter step A filter coefficient power calculating step for calculating a coefficient power sum of the filter coefficient power and a convergence degree determining step for determining a degree of convergence of the filter step based on a calculation result of the filter coefficient power calculating step. This is useful for determining the degree of convergence by directly detecting the update progress of the filter coefficient, which is the coefficient update result. Therefore, it is possible to stably determine the convergence level of the adaptive equalizer even for a signal whose power and spectrum fluctuate from moment to moment such as voice, and noise. Even in the presence, it has the effect of being able to stably determine the degree of convergence.

  In a ninth aspect of the invention made to solve the above-described problem, the convergence determination step detects the amount of change in the coefficient power sum calculated by the filter coefficient power calculation step. Because it is possible to use the update progress of the filter coefficient itself that is not easily affected by fluctuations even in the presence of signals and noise whose power and spectrum fluctuate every moment, it is possible to judge the degree of convergence. It has the effect | action and effect that can be performed stably.

  In a tenth aspect of the invention made to solve the above-described problem, the convergence determination step includes the power sum at the previous determination and the power at the current determination as the amount of change in the coefficient power sum calculated by the filter coefficient power calculation step. The ratio of the sum is used, and it is possible to flexibly respond to two conflicting demands: a demand for reducing the circuit or computation scale and speeding up judgment, and a demand for improving judgment accuracy. It has the action and effect of becoming.

  In an eleventh aspect of the invention made to solve the above-described problem, the convergence determination step includes the power sum at the previous determination and the power at the current determination as a change amount of the coefficient power sum calculated by the filter coefficient power calculation step. The difference between the sums is used, and it is possible to flexibly respond to two conflicting demands: a demand for reducing the circuit or computation scale and speeding up judgment, and a demand for improving judgment accuracy. It has the action and effect of becoming.

  In a twelfth aspect of the invention made to solve the above problem, the convergence determination step calculates a difference between the power sum at the previous determination and the power sum at the current determination, and calculates a change in the sign state of the calculation result. Since it can be determined that the filter is in a converged state when a positive sign and a negative sign appear with equal probability in the sign of the calculation result, the time change of the power sum is simply It has an operation and effect that it is possible to determine the completion of convergence more simply than the determination and the degree of convergence is continuously traced.

  According to a thirteenth aspect of the present invention for solving the above-described problems, the filter coefficient power calculation step is performed by replacing each filter coefficient value with a square calculation in each filter coefficient value power calculation. The absolute value is used, and there is an action and an effect that it becomes possible to cope with a demand for reduction in circuit or calculation scale and speeding up determination.

  In a fourteenth aspect of the invention made to solve the above-described problem, the degree of convergence determination step includes a change amount of the power sum calculated by the filter coefficient power calculation step in synchronization with a timing at which the filter coefficient update step updates the filter coefficient. Stable by preventing the wrong convergence from being judged as no change in power sum is recognized under the timing when the filter update operation is not operating properly It has the effect | action and effect that determination can be performed.

  A fifteenth aspect of the invention made to solve the above-mentioned problems is a computer-readable recording of a program for executing each step of the adaptive equalization method according to any one of the eighth to fourteenth aspects of the invention. The recording medium is a possible recording medium, and the adaptive equalization method according to any of the eighth to fourteenth aspects of the invention can be executed at any time and at any place.・ Has an effect.

(Embodiment 1)
FIG. 1 is a general block diagram for realizing an adaptive equalizer.

  In FIG. 1, 101 is an A / D converter that converts a received voice signal, which is an analog voice signal, into a digital signal and outputs it as a digital voice signal (equalized input signal), and 102 is a digital voice signal (equalized output signal). D / A converter 103 for converting the signal into an analog audio signal 103 performs digital signal processing on the digital audio signal obtained from the A / D converter 101 and outputs the calculation result to the D / A converter 102 A central processing unit (adaptive equalizer), 104 is a ROM (Read Only Memory) in which a program for operating the central processing unit 103 as an echo cancellation device is stored, and 105 is a program stored in the ROM 104. RAM (Random Access) used when the central processing unit 103 operates according to Memory).

  FIG. 2 is a functional block diagram showing the central processing unit 103 as the adaptive equalizer of FIG. 1 and shows functional blocks realized by executing a program stored in the ROM 104.

  In FIG. 2, 201 is a reference signal, 202 is a desired signal (equalized input signal), 203 is an error signal (equalized output signal), 204 is a filter means for performing a convolution operation between the estimated transfer function and the reference signal 201, 205 is a filter output signal of the filter means 204, 206 is a subtraction means for subtracting the filter output signal 205 from the desired signal (equalization input signal) 202 to generate an error signal (equalization output signal) 203, and 207 is a LMS (Least) Transfer function estimation means for estimating a transfer function by means of a Mean Square method, an NLMS (Normalized LMS) method, etc., and 208 a filter coefficient update for reflecting and updating the estimation result of the transfer function estimation means 207 to the filter coefficient in the filter means 204 Means 209 designates a specified group of filter coefficients in the filter means 204. Filter coefficient power calculation means 210 for calculating the coefficient power sum of several or all filter coefficients, and a convergence degree determination means 210 for referring to the output of the filter coefficient power calculation means 209 to determine the degree of convergence of the coefficients of the filter means 204 is there. With this configuration, the degree of convergence of the coefficients of the filter unit 204 can be determined.

  FIG. 3 is a graph comparing the effects of the presence or absence of noise on the adaptation state of the filter coefficient in the filter means 204, and FIG. 3 (a) shows the time of the filter coefficient when the adaptation operation proceeds normally and reaches convergence. FIG. 3B is a graph showing an example of change, and FIG. 3B is a graph showing an example of time change of the filter coefficient when divergence occurs in the middle of the adaptive operation.

  First, as shown in FIG. 3A, a graph showing an example of time variation of filter coefficients when the adaptive operation is normally progressed and converges, as the adaptive operation time elapses, each filter coefficient value gradually increases. It can be read that the operation is stopped when the predetermined time elapses and the change stops and converges. On the other hand, as shown in FIG. 3 (b), a graph showing an example of the time change of the filter coefficient when the divergence is reached in the middle of the adaptive operation, as shown in FIG. It can be seen that the coefficient value suddenly changes again and diverges. Focusing on this filter coefficient fluctuation characteristic, it is possible to determine the degree of convergence of the filter means 204. Therefore, in the present invention, in order to trace the operation of the coefficients of the filter means 204 with a simple configuration, the filter coefficient power for calculating the coefficient power sum of the specified arbitrary number or all the filter coefficients in the filter coefficient group. Arithmetic means 209 is provided.

  Hereinafter, the principle of convergence detection according to the present invention will be described with reference to FIG.

  FIG. 4 shows an example of the time variation of the filter coefficient with respect to FIGS. 3A and 3B in the case where the filter coefficient power calculation unit 209 sets all the filter coefficients as the calculation target for outputting the coefficient power sum. It is a graph which shows the time change of a corresponding coefficient electric power sum. Example of time variation of filter coefficient when adaptive operation is normally advanced and converges With respect to the graph corresponding to FIG. 3A, the power sum converges to a constant value with the elapsed time of the adaptive operation, whereas the adaptive operation FIG. 3B is a graph showing an example of the time change of the filter coefficient when the divergence is caused in the middle of the graph. It can be seen that the power sum suddenly increases or decreases in the middle of the adaptive operation. Therefore, it is possible to determine the convergence of the filter unit 204 by detecting the amount of change in the power sum that is the output of the filter coefficient power calculation unit 209 each time. In FIG. 4, the calculation target in the filter coefficient power calculation unit 209 is all filter coefficients as described above. However, as a method of selecting a filter coefficient to be calculated, the maximum or minimum coefficient is used. A method of continuously selecting an area around a filter coefficient number, a method of selecting only even numbers or odd numbers of filters, and dividing the entire filter into n (n is an integer of 2 or more), and each divided section Various methods are conceivable, such as a method targeting a filter coefficient existing at the center of the filter. In addition, the sum of the squares of the filter coefficients is used for the calculation of the power sum in the filter coefficient power calculation unit 209, but it can be replaced with an expression for calculating another physical quantity in accordance with the power sum of the filter coefficients. .

  Subsequently, the convergence determination operation will be described with reference to FIG.

  FIG. 5 is a flowchart showing an example of the implementation of the convergence degree determination unit 210 of FIG. In FIG. 5, first, when the convergence determination unit 210 starts operation, the latest power sum Sc is acquired from the filter coefficient power calculation unit 209 (S501), and the ratio R with the power sum Sp that has been acquired last time is calculated. Then, it is determined whether or not the ratio is in the vicinity of 1 (S502). Here, either R = Sc / Sp or R = Sp / Sc may be used. In determining whether R is in the vicinity of 1, for example, the lower limit threshold T1 and the upper limit threshold T2 are independent of each other. If there is a method for determining that it exists in T1 <R <T2 (where T1 <1 and T2> 1), A% that is an allowable range for 1 is determined and 1 ± A For example, there is a method for judging whether or not R is present within%. Furthermore, if division is used to determine the ratio, a division by zero (zero division) error may occur. Therefore, either the zero division is handled as an exception, or evaluation is performed using an equivalent expression so that division does not occur (for example, When evaluating whether or not R = Sc / Sp <T2, it is effective to modify the equation to evaluate whether Sc <Sp × T2).

  If it is determined in S502 that the ratio R is close to 1, the value of the counter is increased (S503), otherwise the value of the counter is decreased (S504). Here, the above S502 to S504 are referred to as counter update (S510). Further, it is determined whether or not the counter value has reached a predetermined upper limit value (S505), and if it has reached the upper limit value, it is determined that the convergence has advanced one step and the counter is initialized ( S506). On the other hand, if it is determined in S505 that the value of the counter has not reached the predetermined upper limit value, it is determined whether or not the counter has reached a predetermined lower limit value (S507). If the value has been reached, it is determined that the divergence has advanced by one step, and the counter is initialized (S508). Here, the above S505 to S508 are referred to as a convergence determination step (S511).

  Finally, an update process for overwriting the previous power sum Sp with the latest power sum Sc is performed (S509), and the process ends.

  As the operation of the convergence determination means 210, a simpler method for determining whether or not the convergence has progressed one step at a time without providing a counter as described in S501 to S511 is considered. It is done. An example of this will be described with reference to FIG.

  FIG. 6 is a flowchart showing an example of the implementation of the convergence degree determination unit 210 of FIG. In FIG. 6, first, when the convergence determination unit 210 starts operation, the latest power sum Sc is acquired from the filter coefficient power calculation unit 209 (S601), and the ratio R with the power sum Sp that has been acquired last time is calculated. Then, it is determined whether the ratio is in the vicinity of 1 (S602). In S602, when it is determined that the ratio R is close to 1, it is determined that the convergence has advanced one step (S603), and otherwise, it is determined that the divergence has advanced one step (S604). . Finally, an update process for overwriting the previous power sum Sp with the latest power sum Sc is performed (S605), and the process ends.

  As described above, in FIGS. 5 and 6, as an example of the determination result of the degree of convergence, an example of classification into two values indicating whether the convergence has progressed one step or whether the divergence has advanced one step has been shown. The classification can be similarly performed. In addition, when only the convergence side determination is required, the step related to the diverging side determination can be omitted. Conversely, when only the diverging side determination is required, the step related to the convergence side determination is omitted. It is also easy to do.

  As described above, according to the present embodiment, in an adaptive equalizer used in a communication device, an audio signal processing device, or the like, the transfer function estimation unit 207 that estimates the transfer function to be equalized and the equalization processing are performed. Filter means 204 for performing, filter coefficient updating means 207 for updating the coefficient value of the filter means 204 based on the estimation result in the transfer function estimating means 207, and a specified arbitrary number or all of the filter coefficient groups of the filter means 204 Filter coefficient power calculation means 209 for calculating the coefficient power sum of the filter coefficients of the filter coefficient, and convergence degree determination means 210 for determining the degree of convergence of the filter means 204 based on the calculation result of the filter coefficient power calculation means 209. Thus, the progress of the filter coefficient update is directly detected by the filter coefficient power calculation means 209, Can be used for the determination of the convergence determination means 210, so that the convergence level of the adaptive equalizer can be adjusted even for a signal whose power and spectrum fluctuate every moment such as speech. The determination can be performed stably, and further, the convergence and the determination of the degree of convergence can be performed stably even in the presence of noise.

  Further, the convergence determination means 210 determines the convergence by detecting the amount of change in the coefficient power sum calculated by the filter coefficient power calculation means 209, so that the convergence degree is determined momentarily as in voice. Since it is possible to use the update progress of the filter coefficient itself, which is not easily affected by fluctuations in the presence of signals and noise whose power and spectrum fluctuate, in the convergence determination, the determination of the convergence degree is stable. Has the effect of being able to be performed.

  In addition, the convergence determination unit 210 uses the ratio of the power sum at the previous determination and the power sum at the current determination as the change amount of the coefficient power sum calculated by the filter coefficient power calculation unit 209. There is an effect that it is possible to flexibly cope with two conflicting demands, that is, a demand for reducing the circuit or computation scale and speeding up judgment and a demand for improving judgment accuracy.

  Furthermore, if a computer-readable recording medium in which a program for executing the steps of FIGS. 5 to 6 is recorded is used, the recording medium is read by a computer, and the recording medium shown in FIGS. The method of FIG. 6 can be performed.

(Embodiment 2)
The configuration of the adaptive equalizer according to the second embodiment of the present invention is the same as that of the first embodiment as shown in FIG. The difference of the sum is used for determination. That is, the determination method of the convergence determination unit 210 is different from that of the first embodiment.

  FIG. 7A is a graph showing the change over time in the ratio of the power sum at the previous determination and the power sum at the current determination in the example of FIGS. 3A and 3B, and the convergence of the filter means 204. As the degree increases, the ratio R of the power sum converges in the vicinity of 1, and the operation of the convergence determination unit 210 using this characteristic is as described in the first embodiment.

  On the other hand, FIG.7 (b) is a graph which shows the time change of the fluctuation | variation of the difference of the electric power sum at the time of last determination, and the electric power sum at the time of this time determination in the example of Fig.3 (a) (b). From this graph, it is clear that there is a property that the difference of the power sum converges to near 0 when the convergence of the filter unit 204 becomes high, and it is possible to determine the convergence of the filter unit 204 using this property. It is.

  The operation of the adaptive equalizer according to the present embodiment is the same as the operation shown in the flowchart of FIG. 5 except that the convergence determination means 210 determines the ratio of the power sum at the previous determination and the power sum at the current determination. Instead, the difference between the power sum at the previous determination and the power sum at the current determination is handled.

  FIG. 8 is a flowchart showing an example of the implementation of the convergence degree determination unit 210 of FIG. In FIG. 8, first, when the convergence determination unit 210 starts operation, the latest power sum Sc is acquired from the filter coefficient power calculation unit 209 (S801), and the difference D from the power sum Sp that has already been acquired is calculated. Then, it is determined whether or not the ratio is in the vicinity of 0 (S802). Here, either D = Sc-Sp or D = Sp-Sc may be used. In determining whether D is in the vicinity of 0, for example, the lower limit threshold T1 and the upper limit threshold T2 are independent of each other. If there is a method for determining that the current value exists at T3 <D <T4 (where T3 <0 and T4> 0), the absolute value of D is an arbitrary threshold value T5 (where T5> 0 and 0). There is a method for determining whether or not the value is less than an arbitrary value in the vicinity.

  If it is determined in S802 that the difference D is close to 0, the value of the counter is increased (S803), and if not, the value of the counter is decreased (S804). Here, the above S802 to S804 are referred to as a counter update step (S810).

  Subsequently, it is determined whether or not the counter value has reached a predetermined upper limit value (S805). If the counter value has reached the upper limit value, it is determined that the convergence has advanced one step and the counter is initialized. (S806). On the other hand, if it is determined in S805 that the counter value has not reached the predetermined upper limit value, it is determined whether the counter has reached a predetermined lower limit value (S807). If the value has been reached, it is determined that the divergence has advanced by one step, and the counter is initialized (S808). Here, the above S805 to S808 are referred to as a convergence determination step (S811).

  Finally, an update process for overwriting the previous power sum Sp with the latest power sum Sc is performed (S809), and the process ends.

  As the operation of the convergence determination means 210, a simpler method for determining whether or not the convergence has progressed one step at a time without providing a counter as described in S801 to S811 is considered. It is done. Furthermore, as an example of the determination of the degree of convergence, an example of classification into binary values indicating whether the convergence has advanced one step or whether the divergence has advanced one step has been shown. Is possible. In addition, when only the convergence side determination is required, the step related to the diverging side determination can be omitted. Conversely, when only the diverging side determination is required, the step related to the convergence side determination is omitted. It is also easy to do.

  As described above, according to the present embodiment, the convergence determination unit 210 uses the power sum at the previous determination and the power sum at the current determination as the amount of change in the coefficient power sum calculated by the filter coefficient power calculation unit 209. By using this difference, it becomes possible to flexibly respond to two conflicting demands: a demand for reducing the circuit or computation scale and speeding up judgment, and a demand for improving judgment accuracy. It has the action and effect.

  Furthermore, if a computer-readable recording medium that records a program for executing each step of FIG. 8 is used, the method of FIG. 8 is executed at an arbitrary place at an arbitrary time by reading the recording medium with a computer. can do.

(Embodiment 3)
The configuration of the adaptive equalizer according to the third embodiment of the present invention is the same as that of the second embodiment as shown in FIG. A difference from the power sum is calculated, and a change in the sign state of the calculation result is detected. That is, the determination method of the convergence determination unit 210 is different from that of the second embodiment.

  FIG. 9 is a flowchart showing an example of the implementation of the convergence degree determination unit 210 of FIG. In FIG. 9, first, when the convergence determination means 210 starts operation, the latest power sum Sc is acquired from the filter coefficient power calculation means 209 (S901), and the difference D from the power sum Sp that has already been acquired is calculated. Then, it is determined whether the sign is positive or negative (S902). When 0 is detected in step S902, it is desirable not to classify as either positive or negative, but a method of classifying into either or both can also be selected. In addition, either D = Sc-Sp or D = Sp-Sc may be used for the calculation of D.

  In S902, if it is determined that the sign of the difference D is positive, the value of the positive sign detection counter is increased (S903), and if it is determined that the sign of the difference D is negative, it is negative. The value of the code detection counter is increased (S904). Here, the above S902 to S904 are referred to as a counter update step (S908).

  Subsequently, it is determined whether the ratio of the positive sign detection counter to the negative sign detection counter is close to 1 (S905). In the determination of whether or not the counter ratio in S905 is close to 1, the same method as that described in the first embodiment can be used. Further, if it is determined in S905 that the ratio of the counter is close to 1, it is determined that the convergence has advanced by one step, and the counter is initialized (S906). Here, the above S905 to S906 are referred to as a convergence determination step (S909).

  Finally, an update process is performed to overwrite the previous power sum Sp with the latest power sum Sc (S907), and the process ends.

  As described above, according to the present embodiment, the filter coefficient power calculation unit 209 calculates the difference between the power sum at the previous determination and the power sum at the current determination, and detects a change in the sign state of the calculation result. By doing so, it is possible to determine that the filter is in a converged state when the positive sign and the negative sign appear with equal probability in the sign of the operation result. It has an operation and effect that it is possible to easily determine the completion of convergence rather than continuing to trace the degree of convergence.

  Furthermore, if a computer-readable recording medium recording a program for executing each step of FIG. 9 is used, the method of FIG. 9 is executed at an arbitrary place at an arbitrary time by reading the recording medium with a computer. can do.

(Embodiment 4)
The configuration of the adaptive equalizer according to the fourth embodiment of the present invention is the same as that of the first embodiment as shown in FIGS. 2 and 5, but the filter coefficient power calculation means 209 performs power calculation of each filter coefficient value. Instead of performing the square calculation of each filter coefficient value, the absolute value of each filter coefficient value is used. That is, the calculation formula in the power calculation of the filter coefficient power calculation means 209 is different from that of the first embodiment.

  As the number of coefficients to be calculated by the filter coefficient power calculation means 209 increases and the convergence of the filter means 204 progresses, the value handled by the filter coefficient power calculation means 209 increases and a large amount of calculation and memory are required. To do. Therefore, in the filter coefficient power calculation means 209, instead of calculating the sum of squares of the filter coefficient value, the absolute value sum of the filter coefficient values is calculated, so that an equivalent determination can be performed with less calculation amount and memory. Is possible.

  As described above, according to the present embodiment, in the filter coefficient power calculation means 209, in the power calculation of each filter coefficient value, instead of performing the square calculation of each filter coefficient value, the absolute value of each filter coefficient value By using, it is possible to meet the demands for reducing the circuit or computation scale and speeding up the determination.

(Embodiment 5)
The configuration of the adaptive equalizer according to the fifth embodiment of the present invention is the same as that of the first embodiment shown in FIGS. 2 and 5. The change amount of the power sum calculated by the filter coefficient power calculation means 209 is detected in synchronization with the update timing. That is, it differs from the first embodiment in that the determination timing of the convergence determination unit 210 is more specifically limited.

  The determination timing of the convergence determination means 210 when the adaptive equalizer of the present invention is applied to an audio signal processing apparatus such as an echo canceller will be described with reference to FIG.

  FIG. 10 is a graph showing the operation timing of the filter coefficient updating unit 208 in an audio signal processing device such as an echo canceller. In the audio signal processing apparatus, generally, the timing at which the filter coefficient updating unit 208 of the adaptive equalizer updates the filter coefficient in the filter unit 204 is synchronized with the sound section of the audio signal used as the reference signal 202. In this case, the filter coefficient update operation timing of the filter coefficient update means 208 for the reference signal 202 as shown in FIG. 10A is as shown in FIG. On the contrary, in the silent section of the reference signal voice shown in FIG. 10A, the update operation of the filter coefficient is stopped. Therefore, the convergence degree determination unit 210 performs the filter coefficient power calculation unit 209 during the update operation stop period. When an attempt is made to detect the amount of change in the power sum calculated by (1), an incorrect convergence determination is performed because no change in the power sum is recognized. In order to prevent this erroneous determination, in the present invention, the filter coefficient update unit 208 updates the filter coefficient when the convergence degree determination unit 210 detects the change amount of the power sum calculated by the filter coefficient power calculation unit 209. It shall synchronize with the timing to perform.

  As an example of the synchronization timing, for example, a counter that starts from the start of the filter coefficient update period as shown in FIG. 10C is provided, and a determination is performed by synchronizing with the counter. As described above, various methods such as a method of performing determination at the end point of the filter coefficient update period are conceivable. In any case, a certain period is always required between the timings when the convergence determination unit 210 performs the convergence determination. The requirement in the fifth embodiment is that the above filter coefficient update period exists.

  In the example of FIG. 10, an example is shown in which the filter coefficient update operation stage of the filter coefficient update unit 208 is two stages of update or non-update. However, the update stage includes a plurality of stages exceeding two stages. There may be some cases. In this case, there are a method of interpreting as two steps by providing a threshold value for the step size, and various methods such as adaptively changing the synchronization timing interval according to the size of the update step can be considered.

  Hereinafter, the determination timing synchronization operation of the convergence determination unit 210 shown in the example of FIG. 10C will be described with reference to FIG.

  FIG. 11 is a flowchart illustrating an example of the implementation of the convergence degree determination unit 210 of FIG. In FIG. 11, when the convergence determination unit 210 starts operating, it is first determined whether or not the value of the synchronization counter has reached 0 (S1101). If the counter has reached 0, the convergence determination process is performed. (S1102), the counter is initialized (S1104), and the process ends. Here, S1102 can use the method illustrated in Embodiments 1 to 4 as it is. On the other hand, if it is determined in S1101 that the synchronization counter has not reached 0, then it is determined whether or not the filter coefficient update means 208 is in the filter coefficient update period (S1103). If it is determined, the synchronization counter is initialized (S1104) and the process ends. If it is determined in S1103 that it is in the filter coefficient update period, the synchronization counter is subtracted (S1106), and the process ends.

  The operation start timing of the flowchart which is an example of the implementation shown in FIG. 11 is periodically performed as an appropriate timing according to the system. For example, in the central processing unit (adaptive equalizer) 103 Various methods are conceivable, such as performing in synchronization with the sampling frequency period of the A / D conversion apparatus 101 or D / A conversion apparatus 102, in synchronization with a timer period separately provided in the system. In this flowchart, an example in which the determination timing synchronization is achieved in the convergence determination unit 210 is shown, but various other methods such as a method of determining the determination timing in the filter coefficient update unit 208 are conceivable.

  As described above, according to the present embodiment, the amount of change in the power sum calculated by the filter coefficient power calculation unit 209 in the convergence degree determination unit 210 in synchronization with the timing at which the filter coefficient update unit 208 updates the filter coefficient. By detecting this, it is possible to make a stable determination by preventing an erroneous convergence determination from being made assuming that there is no change in the power sum under the timing when the filter update operation is not operating properly It has the action and effect that it becomes possible to perform.

  Further, if a computer-readable recording medium in which a program for executing each step of FIG. 11 is recorded is used, the method of FIG. 11 is executed at an arbitrary place at an arbitrary time by reading the recording medium with a computer. can do.

The present invention relates to an adaptive equalizer, an adaptive equalization method, and a recording medium for executing the adaptive equalization method used in a communication device, an audio signal processing apparatus, and the like. The convergence degree can be determined stably even for a signal whose spectrum varies, and the convergence state can be determined stably even in the presence of noise.

General block diagram for implementing an adaptive equalizer Functional block diagram showing a central processing unit as the adaptive equalizer of FIG. (A) A graph showing an example of time change of a filter coefficient when the adaptive operation is normally progressed and converged, and (b) a graph showing an example of time change of the filter coefficient when the divergence is reached in the middle of the adaptive operation. Graph showing the time change of the coefficient power sum corresponding to the time change example of the filter coefficient with respect to FIG. 3A and FIG. The flowchart which shows operation | movement of the convergence determination means by Embodiment 1 of this invention. The flowchart which shows operation | movement of the convergence determination means by Embodiment 1 of this invention. (A) In the examples of FIGS. 3 (a) and 3 (b), a graph showing the change over time of the ratio between the power sum at the previous determination and the power sum at the current determination, and (b) FIGS. 3 (a) and 3 (b). In the example, the graph showing the change over time of the difference between the power sum at the previous judgment and the power sum at the current judgment The flowchart which shows the operation | movement of the convergence determination means by Embodiment 2 of this invention. The flowchart which shows operation | movement of the convergence determination means by Embodiment 3 of this invention. Illustration of judgment timing The flowchart which shows the operation | movement of the convergence determination means by Embodiment 5 of this invention. Configuration diagram of basic adaptive equalizer (A) The figure which shows the structural example at the time of applying an adaptive equalizer as ANC (active noise control) with respect to duct noise, (b) The structural example at the time of applying an adaptive equalizer as a line echo canceller in communication equipment Illustration Block diagram showing a conventional adaptive equalizer Functional block diagram showing convergence determination means and power calculation unit of an adaptive equalizer according to the prior art

Explanation of symbols

101 A / D converter 102 D / A converter 103 Central processing unit (adaptive equalizer)
104 ROM
105 RAM
201 Reference signal 202 Desired signal (equalization input signal)
203 Error signal (equalized output signal)
204 Filter means 205 Filter output signal 206 Subtraction means 207 Transfer function estimation means 208 Filter coefficient update means 209 Filter coefficient power calculation means 210 Convergence determination means 1201 Reference signal 1202 Unknown system 1203 Adaptive equalizer 1204 Desired signal 1205 Filter output signal 1301 Duct 1302 Duct noise path (unknown system)
1303 First microphone 1304 Phase inverter 1305 Speaker 1306 Second microphone 1307 Active noise control device 1308 Line echo path 1309 Echo canceller 1401 Adder 1402 Adaptive filter 1403 Echo canceller unit 1404 Hybrid circuit 1405 Power calculation unit 1406 Convergence degree determination unit 1407 Power calculation unit 1408 Delay circuit 1409 Reference signal 1410 Input signal 1411 Output signal 1501 Multiplier 1502 Coefficient unit 1503 Adder 1504 Coefficient unit 1505 Delay circuit 1506 Divider 1507 Comparison determination control unit 1508 Threshold setting unit

Claims (15)

An adaptive equalizer used in a communication device, an audio signal processing device, or the like, in which a transfer function estimation unit that estimates a transfer function to be equalized, a filter unit that performs equalization processing, and the transfer function estimation unit Filter coefficient updating means for updating the coefficient value of the filter means based on the estimation result, and filter coefficient power calculating means for calculating the coefficient power sum of a specified arbitrary number or all filter coefficients in the filter coefficient group of the filter means And an convergence equalizer for determining a degree of convergence of the filter unit based on a calculation result of the filter coefficient power calculation unit. The adaptive equalizer according to claim 1, wherein the convergence determination unit detects a change amount of the coefficient power sum calculated by the filter coefficient power calculation unit. 3. The convergence determination means uses a ratio of the power sum at the previous determination and the power sum at the current determination as the amount of change in the coefficient power sum calculated by the filter coefficient power calculation means. An adaptive equalizer described in 1. 3. The convergence determination unit uses a difference between the power sum at the previous determination and the power sum at the current determination as the amount of change in the coefficient power sum calculated by the filter coefficient power calculation unit. An adaptive equalizer described in 1. The adaptation according to claim 4, wherein the convergence determination means calculates a difference between a power sum at the previous determination and a power sum at the current determination, and detects a change in a sign state of the calculation result. Equalizer. The filter coefficient power calculation means uses an absolute value of each filter coefficient value instead of performing a square calculation of each filter coefficient value in the power calculation of each filter coefficient value. Adaptive equalizer. 3. The convergence determination unit detects a change amount of the power sum calculated by the filter coefficient power calculation unit in synchronization with a timing at which the filter coefficient update unit updates the filter coefficient. The adaptive equalizer described. An adaptive equalization method used in a communication device, an audio signal processing device, or the like, wherein a transfer function estimation step for estimating a transfer function to be equalized, a filter step for performing equalization processing, and the transfer function estimation step A filter coefficient updating step for updating a coefficient value of the filter step based on an estimation result, and a filter coefficient power calculating step for calculating a coefficient power sum of a specified arbitrary number or all of the filter coefficients in the filter coefficient group of the filter step And a convergence determination step for determining a convergence degree of the filter step based on a calculation result of the filter coefficient power calculation step. 9. The adaptive equalization method according to claim 8, wherein the convergence determination step detects a change amount of the coefficient power sum calculated by the filter coefficient power calculation step. 10. The convergence determination step uses a ratio of the power sum at the previous determination and the power sum at the current determination as a change amount of the coefficient power sum calculated by the filter coefficient power calculation step. The adaptive equalization method described in 1. 10. The convergence determination step uses a difference between the power sum at the previous determination and the power sum at the current determination as the amount of change in the coefficient power sum calculated by the filter coefficient power calculation step. The adaptive equalization method described in 1. The adaptation according to claim 11, wherein the convergence determination step calculates a difference between a power sum at the previous determination and a power sum at the current determination, and detects a change in a code state of the calculation result. Equalization method. 9. The filter coefficient power calculation step uses an absolute value of each filter coefficient value instead of performing a square calculation of each filter coefficient value in the power calculation of each filter coefficient value. Adaptive equalization method. 9. The convergence determination step detects a change amount of the power sum calculated by the filter coefficient power calculation step in synchronization with a timing at which the filter coefficient update step updates the filter coefficient. The adaptive equalization method described. The computer-readable recording medium which recorded the program for performing each step of the adaptive equalization method of any one of Claims 8 thru | or 14.

Images