JP2006311356A - Howling detecting device and method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To suppress howling before the howling can be confirmed in an auditory manner by detecting the howling only with arithmetic processing without depending upon the signal level of an input signal. <P>SOLUTION: The howling detecting device 1'' comprises an input terminal 6 for receiving an input signal to be a target for detecting howling, an input part 2 for converting an analog input signal input from the input terminal 6 into a digital one, a signal sectioning part 3 for sampling a finite length section of a signal output from the input part 2, a complex frequency calculating part 4 for calculating a complex frequency of the sampled signal, a howling determining part 5 for detecting the existence/absence of howling on the basis of a code of an imaginary part of the complex frequency, and an output terminal 7 for outputting a result of the existence/absence of howling. When the code of the imaginary part is positive, generation of howling is detected. The detected howling is controlled by a howling control part 9. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a technique for detecting howling in an electric circuit that handles analog signals.

  2. Description of the Related Art Conventionally, an electric circuit has been used that is configured to convert physical amplitudes such as voice and music into electrical signals and collect them, amplify the collected electrical signals, and output them again as physical amplitudes. Among these electric circuits, a technique for detecting unnecessary howling in an electric circuit that handles analog signals has been proposed.

  As a howling detection technique, for example, a signal level that increases with howling in an input signal is detected as an increase in sound volume due to howling, or an increasing signal level is detected electrically as a howling component.

As another howling detection technique, in order to discriminate between a normal frequency characteristic that does not include a howling component and a frequency characteristic that includes a howling component in a howling state, the power frequency characteristic is differentiated on the frequency axis. A technique that uses the obtained value for howling detection is disclosed (see Patent Document 1).
JP 7-147700 A

  However, according to the conventional technique, depending on the increase in the howling signal level in the input signal, howling is audibly confirmed and howling is detected, or the power spectrum of the electric signal at the signal level at which howling has occurred, It is necessary to detect the howling by analyzing the power envelope and so on, and therefore the howling cannot be detected unless the howling signal level has increased to a level at which the howling causes audible discomfort. .

  The present invention has been made in view of the above problems, and an object of the present invention is to detect howling before the feedback reaches a signal level that can be audibly confirmed without depending on the signal level of howling in the input signal. It is in.

  In order to solve the problem, the present invention according to claim 1 includes an input unit that converts an analog input signal into digital, and a signal section that samples a finite-length section from a time series of the input signal converted into digital. And a complex frequency calculation unit for calculating one or more complex frequencies F = f + ib that characterize the time-series signal in this section for each section of the time series sampled by the signal segmentation unit, and the calculated complex frequency A howling detection unit that detects the presence or absence of howling based on the sign of the imaginary part b of F = f + ib.

  In the present invention, the complex frequency calculation unit performs only the calculation processing of the complex frequency F = f + ib on the time series signal sampled from the time series without depending on the howling signal level in the input signal, and the imaginary part of the calculation result Based on the sign b, the occurrence of howling is detected before the howling reaches a signal level that can be audibly confirmed.

According to a second aspect of the present invention, in the first aspect of the present invention, the signal partitioning unit includes K complex frequency F _k = f _k in the output time series signal s (t). Operator with + ib _k as parameter

The calculation of the complex frequency F _k = f _k + ib _k that becomes a random time series is executed.

According to the present invention, the operator according to claim 2 is operated with the complex frequency F _k = f _k + ib _k as a parameter in the calculation of the complex frequency F = f + ib without depending on the howling signal level in the input signal. The complex frequency F _k = f _k + ib _k is calculated so that the time series obtained in this way becomes a random time series, and howling is performed before the howling becomes a signal level that can be audibly confirmed based on the sign of the imaginary part b. Detect the occurrence of

According to a third aspect of the present invention, in the first aspect, the signal segmentation unit includes K complex frequency F _k = f _k in the output time series signal s (t). Operator with + ib _k as parameter

The calculation of the complex frequency F _k = f _k + ib _k that minimizes the autocorrelation is performed.

In the present invention, the operator according to claim 3 is operated with the complex frequency F _k = f _k + ib _k as a parameter in the calculation of the complex frequency F = f + ib without depending on the howling signal level in the input signal. The complex frequency F _k = f _k + ib _k that minimizes the time-series autocorrelation obtained in this way is calculated, and howling is performed before the howling becomes a signal level that can be audibly confirmed based on the sign of the imaginary part b. Detect the occurrence of

According to a fourth aspect of the present invention, in the first aspect of the present invention, the signal partitioning unit includes K complex frequency F _k = f _k in the output time series signal s (t). Operator with + ib _k as parameter

The calculation of the complex frequency F _k = f _k + ib _k that minimizes is performed.

In the present invention, the operator according to claim 4 is obtained by calculating the complex frequency F _k = f _k + ib _k as a parameter in the calculation of the complex frequency F = f + ib without depending on the signal level of the input signal. The complex frequency F _k = f _k + ib _k that minimizes the second-order norm of the time series to be calculated is calculated, and howling occurs before the howling becomes a signal level that can be audibly confirmed based on the sign of the imaginary part b. Is detected.

According to a fifth aspect of the present invention, in the first aspect, the K complex frequencies (K is set to 1 or more) using the linear prediction method from the time series signal s (t) output by the signal partitioning unit. to extract the _{F k = f k + ib k} .

In the present invention, the linear prediction method is used to extract K (K is 1 or more) complex frequencies F _k = f _k + ib _k .

According to a sixth aspect of the present invention, in the first aspect, howling occurs when there is one or more complex frequencies in which the sign of the imaginary part b _k of the complex frequency F _k = f _k + ib _k is positive. To detect.

In the present invention, the detection criterion of howling occurs is determined when there is one or more complex frequencies in which the sign of the imaginary part b _k of the complex frequency F _k = f _k + ib _k is positive.

According to a seventh aspect of the present invention, in the first aspect, howling is performed at a frequency corresponding to the real part f _k of the complex frequency in which the sign of the imaginary part b _k of the complex frequency F _k = f _k + ib _k is positive. It is determined that it has occurred.

In the present invention, howling is detected at a frequency corresponding to the real part f _k of the complex frequency _where the sign of the imaginary part b _k of the complex frequency F _k = f _k + ib _k is positive.

The present invention according to claim 8 relates to the frequency corresponding to the real part f _k of the complex frequency in which the sign of the imaginary part b _k of the complex frequency F _k = f _k + ib _k is positive. A howling suppression unit that suppresses howling using a notch filter having suppression characteristics is provided.

  In the present invention, howling that occurs using a notch filter is suppressed before the howling reaches a signal level that can be audibly confirmed.

The present invention according to claim 9 uses the input signal s (t) in a piecewise manner using K complex coefficients a _k and complex frequencies F _k = f _k + ib _k.

And the occurrence of howling is detected based on the sign of the imaginary part b _k for each section.

  In the present invention, the complex frequency F = f + ib is calculated by the complex frequency calculation unit for the time series signal sampled from the time series of the input signal without depending on the howling signal level in the input signal, and the sign of the imaginary part b The occurrence of howling is detected before the howling becomes a signal level that can be audibly confirmed based on the above.

According to a tenth aspect of the present invention, in the ninth aspect, the input signal s (t) is obtained using a least square method.

Approximate with the following waveform.

  In the present invention, the least square method is used to approximate the input signal s (t).

The present invention according to claim 11, in claim 9, howling when those codes of the imaginary part b _k of the complex frequency F _k = f _k + ib _k is positive there is at least one is generated It is determined that

In the present invention, in order to detect howling before the signal level can be audibly confirmed, there is one or more that have a positive sign of the imaginary part b _k of the complex frequency F _k = f _k + ib _k The occurrence of howling is detected when there is a case.

According to a twelfth aspect of the present invention, in the ninth aspect, howling is performed at a frequency corresponding to the real part f _k of the complex frequency in which the sign of the imaginary part b _k of the complex frequency F _k = f _k + ib _k is positive. It is determined that it has occurred.

In the present invention, the real part f of the complex frequency whose sign of the imaginary part b _k of the complex frequency F _k = f _k + ib _k is positive in order to detect howling before the signal level can be audibly confirmed. Howling is detected at a frequency corresponding to _k .

The present invention according to claim 13 relates to the frequency corresponding to the real part f _k of the complex frequency in which the sign of the imaginary part b _k of the complex frequency F _k = f _k + ib _k is positive. A howling suppression unit that suppresses howling using a notch filter having suppression characteristics is provided.

  In the present invention, the howling generated using the notch filter is suppressed before reaching a signal level that can be audibly confirmed without depending on the signal level of howling in the input signal.

The present invention according to a fourteenth aspect relates to an imaginary part calculated in the section, in contrast to the howling suppression for the time-series section used to calculate the complex frequency F _k = f _k + ib _k in the thirteenth aspect. the sign of b _k is added to the real part f _k of positive a complex frequency, the notch also the real part f _k the sign is positive a complex frequency of the imaginary part b _k calculated in zero or more of the section preceding this Used as a filter parameter to suppress howling.

In the present invention, without depending on the howling signal level in the input signal, the howling suppression for the time-series interval used to calculate the complex frequency F _k = f _k + ib _k is calculated in the interval. is the code of the imaginary part b _k is added to the real part f _k of positive a complex frequency, the real part f _k the sign is positive a complex frequency of the imaginary part b _k calculated in zero or more of the section preceding this Is used as a parameter of the notch filter to suppress the generated howling before reaching a signal level that can be audibly confirmed.

  According to the present invention, howling can be detected only by arithmetic processing without depending on the signal level of howling in an input signal.

  Another effect of the present invention can be suppressed before the howling can be audibly confirmed.

<First Embodiment>
FIG. 1 shows a configuration diagram of a howling detection apparatus according to the first embodiment. In FIG. 1, a howling detection device 1, an input terminal 6 provided with an input signal to be detected in howling detection device 1, and an analog input signal input from the input terminal 6 are digitally provided. An input unit 2 that converts to a signal, a signal segmentation unit 3 that samples a finite-length section of the signal output from the input unit 2, and a complex frequency calculation unit that calculates a complex frequency of the sampled signal and obtains an imaginary part 4, a howling determination unit 5 that detects the occurrence of howling based on the sign of an imaginary part, and an output terminal 7 that outputs a result of the presence or absence of howling.

  In such a configuration, the input unit 2 can use a signal conversion circuit such as an A / D converter. And the structure of the signal partitioning part 3, the complex frequency calculation part 4, and the howling determination part 5 is realizable with the structure of the logic circuit by hardware, for example.

  Next, the flow of signal processing for detection of howling by the howling detection apparatus 1 having such a configuration will be described with reference to the flowchart of FIG.

  First, an input signal is input to the input terminal 6. This input signal is a signal to be detected for occurrence of howling, and is an analog signal. The inputted input signal is subjected to A / D conversion signal processing in the input unit 2 to become a digital signal and obtain a time series s (t) (step 1: S1). The sampling frequency of the A / D conversion executed by the input unit 2 is, for example, 40 kHz or more is selected for howling detection up to 20 kHz, and 1 GHz or more is selected for howling detection up to 500 MHz.

Then the n-th segment of the time series s (t) by the signal segmentation section 3 cuts out a time series s (t) between times t _n of the time t _{n + 1} (Step 2: S2). For the partitioning performed by the signal partitioning unit 3, for example, a method of taking t _{n + 1} −t _n so as to be constant is used.

Next, an operator is applied to the time series s (t) cut out by the signal segmentation unit 3 in the complex frequency calculation unit 4.

A complex frequency F _k = f _k + ib _k is calculated such that becomes a random time series (step 3: S3).

As a method for calculating the complex frequency F _k = f _k + ib _k , a time series is used.

For calculating the complex frequency F _k = f _k + ib _k that minimizes the

A method of calculating a complex frequency F _k = f _k + ib _k that minimizes the autocorrelation may be used.

Further, the input signal s (t) is piecewise divided into K pieces (K is 1 or more) using a complex coefficient a _k and a complex frequency F _k = f _k + ib _k.

The complex frequency F _k = f _k + ib _k may be calculated using an approximation method for approximating with the following waveform.

  Any of the above calculation methods can be similarly extracted only by signal processing by calculation without depending on the increase in the howling signal level generated in the input signal.

Next, if at least one of the complex frequencies F _k = f _k + ib _k calculated by the complex frequency calculation unit 4 in the howling determination unit 5 has a positive sign of the imaginary part b _k , this is used. Howling is detected, and a detection result is output to the output terminal 7 (step 4: S4).

  According to the first embodiment described above, howling can be detected only by arithmetic processing without depending on the signal level of howling in the input signal.

<Second Embodiment>
FIG. 3 shows a configuration diagram of a howling detection apparatus according to the second embodiment. The second embodiment is different from the already described first embodiment in that a howling detection apparatus 1 ′ includes a linear prediction coefficient calculation unit 8 and a howling suppression unit 9. In the first embodiment, whether or not howling is included in the input signal is detected and only the detection result is output. In the second embodiment, howling that occurs is suppressed and howling is performed. The signal from which is removed is output.

  FIG. 4 is a flowchart for explaining howling detection and its suppression signal processing executed in the configuration shown in FIG.

  First, the input signal input to the input terminal 6 is A / D converted by the input unit 2 to become a digital signal to obtain a time series s (t) (step 10: S10). As in the first embodiment, the sampling frequency of A / D conversion in the input unit 2 is, for example, 40 kHz or more is selected for howling detection up to 20 kHz, and 1 GHz or more is selected for howling detection up to 50 MHz. Is done.

Next, as the n-th segment of the time series s (t) in the signal segmentation unit 2, the time series s (t) from the time t _n to the time t _{n + 1} and the time series s (t) ′ having the same contents are used. Is cut out (step 11: S11).

Next, the linear prediction coefficient calculation unit 8 calculates K prediction coefficients a _k (K is 1 or more) using the linear prediction method for the time series s (t) cut out by the signal segmentation unit 3. (Step 12: S12).

Next, an equation using the prediction coefficient a _k calculated by the linear prediction coefficient calculation unit 8 in the howling determination unit 5 as a coefficient.

The of the K solution c _k obtained by solving for x, even one | c _k |> 1 becomes solution c _k is detected as howling when present (step 13: S13).

Next, the complex frequency calculation unit 4 calculates the complex frequency F _k = f _k + ib _k corresponding to the howling by taking the logarithm of the solution c _{k such} that | c _k |> 1 (F _k = −ilnc _k ) ( Step 14: S14).

Next, the howling suppression unit 9 forms a notch filter (not shown) that attenuates the frequency corresponding to the real part f _k of the complex frequency F _k = f _k + ib _k corresponding to howling, and is cut out by the signal partitioning unit 3. The time series s (t) ′ is calculated to suppress howling (step 15: S15). The notch filter can be realized by a filtering circuit configured by hardware.

  Next, in the output unit 10, the input signal whose feedback is suppressed by the howling suppression unit 9 is D / A converted and returned to the original analog signal, and then output from the output terminal 7.

  According to the second embodiment described above, howling can be detected only by arithmetic processing without depending on the signal level of howling in the input signal.

  As another effect, howling can be suppressed before it can be audibly confirmed.

<Third Embodiment>
FIG. 5 shows a configuration diagram of a howling detection apparatus according to the third embodiment. The third embodiment is different from the already described first embodiment in that a howling detection device 1 ″ includes a howling suppression unit 9. In the first embodiment, whether or not howling is included in the input signal is detected and only the detection result is output. In the third embodiment, the same as in the second embodiment. The howling which occurred in this is suppressed and the signal which removed howling is output.

  FIG. 6 is a flowchart for explaining howling detection and suppression signal processing executed in the configuration shown in FIG.

  First, the input signal input to the input terminal 6 is A / D converted by the input unit 2 to become a digital signal to obtain a time series s (t) (step 20: S20). As in the first and second embodiments, the sampling frequency of A / D conversion in the input unit 2 is, for example, 40 kHz or more is selected for howling detection up to 20 kHz, and 1 GHz for howling detection up to 50 MHz. The above is selected.

Next, as the n-th segment of the time series s (t) in the signal segmentation unit 2, the time series s (t) from the time t _n to the time t _{n + 1} and the time series s (t) ′ having the same contents are used. Is cut out (step 21: S21).

Next, the complex frequency calculation unit 4 uses the least square method of the time series s (t).

The complex frequency F _k = f _k + ib _k is calculated by approximating with the following waveform (step 22: S22).

The calculation of the complex frequency F _k = f _k + ib _k is, for example, an equation that gives an approximate residual E

It is sufficient to adopt a method such as

Note that operator that the complex frequency F _k = f _k + ib _k of the K time series signals s (t) in the calculation of the complex frequency _{F k = f k + ib k} (K is 1 or more) parameters

Calculation of the complex frequency F _k = f _k + ib _k that becomes a random time series may be executed.

Or an operator whose parameter is K complex frequencies F _k = f _k + ib _k (K is 1 or more) in the time series signal s (t).

The complex frequency F _k = f _k + ib _k may be calculated with the smallest autocorrelation.

Alternatively, an operator whose parameter is K complex frequencies F _k = f _k + ib _k (K is 1 or more) in the time series signal s (t).

Calculation for obtaining the complex frequency F _k = f _k + ib _k that minimizes may be executed. Any of these calculation methods can be similarly extracted by only signal processing by calculation without depending on the increase in the howling signal level generated in the input signal.

Next, in the howling determination unit 5, howling is present when at least one of F _k = f _k + ib _k calculated by the complex frequency calculation unit 4 has a positive sign of the imaginary part b _k. (Step 23: S23).

Next, in the howling suppression unit 9, the complex frequency F _k = f _k + ib _k corresponding to the howling in the main section cut out by the signal partitioning unit 3 and the L sections (L is 0 or more) preceding this section. A notch filter (not shown) that attenuates the frequency corresponding to the real part fk is constructed, and the time series s (t) ′ cut out by the signal partitioning unit 3 is calculated to suppress howling (step 24: S24). .

  Next, the signal whose feedback is suppressed by the howling suppression unit 9 in the output unit 10 is D / A converted and returned to the original analog format input signal, and output from the output terminal 7.

  According to the third embodiment described above, howling can be detected only by arithmetic processing without depending on the signal level of howling in the input signal.

  As another effect, howling can be suppressed before it can be audibly confirmed.

The block diagram of the howling detection apparatus of 1st Embodiment is shown. The flowchart of the howling detection process in 1st Embodiment is shown. The block diagram of the howling detection apparatus of 2nd Embodiment is shown. The flowchart of the howling suppression process in 2nd Embodiment is shown. The block diagram of the howling detection apparatus of 3rd Embodiment is shown. The flowchart of the howling suppression process in 3rd Embodiment is shown.

Explanation of symbols

1, 1 ', 1''Howling suppression device 2 Input unit 3 Signal segmentation unit 7
4 Complex Frequency Calculation Unit 5 Howling Determination Unit 6 Input Terminal 7 Output Terminal 8 Linear Prediction Coefficient Calculation Unit 9 Howling Suppression Unit 10 Output Unit

Claims (14)

An input unit for converting an analog input signal to digital, and
A signal partitioning unit that samples a finite-length section from a time series of the input signal converted into digital;
A complex frequency calculation unit that calculates one or more complex frequencies F = f + ib that characterize the time-series signal in this interval for each time-series interval sampled by the signal partitioning unit;
A howling detection unit for detecting the presence or absence of howling based on the sign of the imaginary part b of the calculated complex frequency F = f + ib;
A howling detection apparatus comprising: In the signal segmentation unit,
An operator whose parameters are K complex frequencies F _k = f _k + ib _k in the output time series signal s (t).

2. The howling detection apparatus according to claim 1, wherein calculation of the complex frequency F _k = f _k + ib _k that becomes a random time series is executed. In the signal segmentation unit,
An operator whose parameters are K complex frequencies F _k = f _k + ib _k in the output time series signal s (t).

2. The howling detection apparatus according to claim 1, wherein the calculation of the complex frequency F _k = f _k + ib _k having the smallest autocorrelation is performed. In the signal segmentation unit,
An operator whose parameters are K complex frequencies F _k = f _k + ib _k in the output time series signal s (t).

2. The howling detection apparatus according to claim 1, wherein the calculation of the complex frequency F _k = f _k + ib _k that minimizes is performed. 3. K complex frequencies F _k = f _k + ib _k are extracted from the time-series signal s (t) output by the signal segmentation unit using a linear prediction method. The howling detection apparatus according to claim 1. The howling detection apparatus according to claim 1, wherein a case where one or more complex frequencies having a positive sign of an imaginary part b _k of the complex frequency F _k = f _k + ib _k is detected as occurrence of howling. . Claim 1, wherein determining that howling in the complex frequency F _k = f _k + ib frequency code of the imaginary part b _k of _k corresponds to the real part f _k of the complex frequency is positive is occurring The howling detection apparatus as described. Howling suppression that suppresses howling using a notch filter having a suppression characteristic for the frequency corresponding to the real part f _k of the complex frequency whose sign of the imaginary part b _k of the complex frequency F _k = f _k + ib _k is positive. The howling detection apparatus according to claim 1, further comprising a unit. The input signal s (t) is divided into K pieces (K is 1 or more) using a complex coefficient a _k and a complex frequency F _k = f _k + ib _k.

A howling detection method characterized by detecting the occurrence of howling based on the sign of an imaginary part b _k for each section. Using the least square method, the input signal s (t)

The howling detection method according to claim 9, wherein the feedback is approximated by the following waveform. The method of claim 9, wherein howling occurs when one or more of the complex frequencies F _k = f _k + ib _k have a positive sign of the imaginary part b _k exist. Howling detection method. 10. It is determined that howling occurs at a frequency corresponding to a real part f _k of a complex frequency whose sign of an imaginary part b _k of the complex frequency F _k = f _k + ib _k is positive. The howling detection method described in 1. Howling suppression that suppresses howling using a notch filter having a suppression characteristic for the frequency corresponding to the real part f _k of the complex frequency whose sign of the imaginary part b _k of the complex frequency F _k = f _k + ib _k is positive. 10. The howling detection method according to claim 9, further comprising a section. In contrast to the howling suppression for the time-series interval used to calculate the complex frequency F _k = f _k + ib _k , the sign of the imaginary part b _k calculated in the interval becomes the real part f _k of the positive complex frequency. In addition, the real part f _k of a complex frequency with a positive sign of the imaginary part b _k calculated in zero or more sections preceding this is also used as a parameter of the notch filter to suppress howling. The howling detection method according to claim 13.

Images