JPH0586718B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION Field of the Invention The present invention relates to a howling suppression device that uses a digital filter to suppress howling in amplified sound.

Conventional Structure and Problems In recent years, with the development of conference systems and teleconference systems, there has been a desire for a howling suppression device that can provide clear and natural sound amplification.

A conventional howling suppression device will be explained below. FIG. 1 is a block diagram of a howling suppression device using a conventional digital filter. 1st
In the figure, 1 is a microphone, 2 is a microphone amplifier, 3 is a subtraction section, 4 is a transfer characteristic estimating section that internally stores transfer characteristics in the form of an impulse response, and 5 is a switch that controls the operation of this transfer characteristic estimating section 4. When this switch 5 is closed, the transfer characteristic estimation operation is performed, and when it is open, the estimation operation is stopped. 6 is a digital filter (hereinafter referred to as echo signal estimator) that performs convolution of the impulse response stored in the transfer characteristic estimator 4 and the speaker amplifier input signal;
7 is a speaker amplifier, 8 is a speaker, 9 is a switch for cutting off the amplification system and inputting white noise to the transfer characteristic estimation section, and 10 is a white noise generation section.

The operation of the conventional howling suppression device configured as described above will be described below.

First, the switch 5 is closed, the switch 9 is connected to the B side, and the transfer characteristic between the speaker 8 and the microphone 1, that is, the impulse response is estimated. The output of the white noise generator 10 is amplified by a speaker amplifier 7 and a speaker 8, and the sound is collected by a microphone 1. The output of the white noise generator 10 is also input to the echo signal estimator 6. The echo signal estimation section 6 calculates the convolution of this white noise signal and the estimated impulse response sent from the transfer characteristic estimation section 4 to obtain an estimated echo signal, which is subtracted from the output of the microphone amplifier 2 by the subtraction section 3. . The output of the subtraction section 3 is sent to the transfer characteristic estimating section 4, and the transfer characteristic estimating section 4 modifies the estimated impulse response stored internally so that the output of the subtraction section 3 decreases. Next, using this revised estimated impulse response, the estimated echo signal is calculated and subtracted again to obtain a new output of the subtractor 3, and the estimated impulse response is again corrected in the direction of decreasing this output. By repeating the above procedure a sufficient number of times, the output of the subtraction unit 3 becomes zero, and as a result, an impulse response with a high degree of approximation is stored inside the transfer characteristic estimation unit 4. After the estimation of the transfer characteristic is completed in this manner, the switch 5 is opened to prohibit modification of the estimated impulse response.

Next, switch 9 is connected to the A side for amplification.
During this loudspeaker operation, the echo signal that circulates from the speaker 8 in the output of the microphone 1 is canceled, the howling loop is blocked, and the occurrence of howling is suppressed.

The calculations of the transfer characteristic estimating section 4 and the echo signal estimating section 6 described above are performed internally digitally. Now, the input signal of the speaker amplifier 7 input to the echo signal estimator 6 is x _(o) , and the estimated impulse response input from the transfer characteristic estimator 4 is
Assuming that h _(o) , the estimated echo signal y is calculated by the convolution operation shown in equation (1).

y= _o=0 〓 ^N x _(-o)・h _(o) ...(1) Here, N is the total number of sampling data obtained by digitizing the analog signals of x and h (hereinafter referred to as the number of taps), n is the number of sampling data.

Assuming that the sampling period for digitization is Δt, the time length T of the estimated impulse response and the high frequency limit F are determined by equation (2).

T=N・Δt F=1/2・Δt} ...(2) Therefore, if the data sampling period is the same, the larger the number of taps, the more long echoes can be removed, and the shorter the sampling period, the higher the frequency. Even the echoes of can be removed.

However, in the configuration of the above embodiment, since it is necessary to remove the echo of the signal up to the high frequency range, the sampling period becomes short, and if the number of taps cannot be large, the reverberation time that can be removed becomes short and sufficient feedback cannot be obtained. No suppression effect can be obtained.
In addition, an increase in the number of taps leads to an increase in the calculation time of equation (1), which poses problems such as making real-time processing impossible and significantly increasing costs.

Purpose of the invention The present invention solves the above-mentioned conventional problems.
It is an object of the present invention to provide a howling suppression device that is low in cost and highly effective in suppressing howling by limiting the echo cancellation band by the echo signal estimator to only low frequencies.

Structure of the Invention The present invention includes a low-pass filter that extracts low-frequency components of the output of a microphone, a high-pass filter that extracts high-frequency components of the output of the microphone, a cutoff frequency adjustment section that adjusts the cutoff frequency of the high-pass filter, and a microphone. a transfer characteristic estimator that estimates a low-frequency transfer characteristic between speakers; a reverberant signal estimator that estimates a low-frequency reverberant signal from the speaker to a microphone; and a low-frequency reverberation estimated from the output of the low-pass filter. It is composed of a subtraction section that subtracts a signal, and an addition section that adds the output of the subtraction section and the output of the high-pass filter.Howling in the low frequency range is suppressed by canceling the echo signal, and howling in the middle frequency range is suppressed. By suppressing this by adjusting the cut-off frequency of a high-pass filter, it is possible to provide a howling suppression device that has a high howling suppression effect and is low in cost.

DESCRIPTION OF EMBODIMENTS FIG. 2 shows a block diagram of a howling suppression device according to a first embodiment of the present invention. In Figure 2, 1 is a microphone, 2 is a microphone amplifier, 21 is a low pass filter (LPF),
22 is a high pass filter (HPF), 23 is a cut-off frequency adjustment section, 3 is a subtraction section, 24 is a transfer characteristic estimation section, and 25 is a switch that controls the operation of this transfer characteristic estimation section 24, and this switch 25 is closed. When the gate is open, the estimation operation of the transfer characteristic is performed, and when the gate is open, the estimation operation is stopped. 26 is a digital filter (hereinafter referred to as echo signal estimation section) that performs convolution of the impulse response stored in the transfer characteristic estimation section 24 and the speaker amplifier input signal; 27 is an addition section; 7 is a speaker amplifier;
8 is a speaker; 9 is a switch for cutting off the amplification system and inputting white noise when estimating transfer characteristics;
10 is a white noise generating section.

The operation of the howling suppression device of this embodiment configured as described above will be described below.

First, the switch 25 is closed, the switch 9 is connected to the B side, and the transfer characteristic between the speaker 8 and the microphone 1 is estimated. White noise generator 1
The output of 0 is amplified by the speaker amplifier 7 and the speaker 8 and collected by the microphone 1. The output of the white noise generator 10 is also input to the echo signal estimator 26. The echo signal estimation section 26 calculates the convolution of this white noise signal and the estimated impulse response sent from the transfer characteristic estimation section 24 to obtain an estimated echo signal, which is subtracted from the output of the low-pass filter 21 by the subtraction section 3. The output of the subtraction section 3 is sent to the transfer characteristic estimation section 24, and the transfer characteristic estimation section 24 modifies the estimated impulse response stored internally in a direction such that the output of the subtraction section 3 decreases. Thereafter, by the same operation as in the conventional example, a low frequency impulse response having a high degree of approximation is stored in the transfer characteristic estimating section 24. After the estimation of the transfer characteristic is completed in this manner, the switch 25 is opened to prohibit modification of the estimated impulse response.

Next, switch 9 is connected to the A side to perform amplification. During amplification, the output of the microphone amplifier 2 is band-divided by a low-pass filter 21 and a high-pass filter 22. The signal containing only low frequency components is processed by the low-pass filter 21, and the echo signal contained therein is removed by the subtraction unit 3, the transfer characteristic estimation unit 24, and the echo signal estimation unit 26. The low frequency component from which this echo signal has been removed is added to the high frequency component separated by the high pass filter 22 in an adder 27 and amplified by a speaker amplifier 7 and a speaker 8. In this loudspeaker system, low-frequency feedback signals are canceled, so no low-frequency howling occurs. Furthermore, howling in high frequencies is also difficult to occur for reasons described later, and howling mainly occurs only in mid-range frequencies. Therefore, howling can be effectively removed by increasing the foot-off frequency of the high-pass filter by the cut-off frequency adjustment section 23 and setting the cut-off frequency above the band where howling occurs.

Next, the principle of the present invention will be explained using actual measured values of transfer characteristics between a speaker and a microphone. Figure 3 shows an example of the actual frequency dependence of the transfer characteristics between a speaker and a microphone, measured in two types of lecture halls with a distance of 1 m between the speakers and microphones. As shown in this example, the values of the transfer characteristics 31 and 32 become smaller as the frequency becomes higher due to the sound absorption effect of walls and the like in any sound field. Therefore, the frequency at which howling occurs is mainly at low frequencies up to 1.5KHz. Furthermore, at frequencies from 1.5KHz to 3KHz, as is clear from the measured values of different sound fields in Figure 3, it is easily influenced by the sound field, and depending on the sound field, the transmission characteristics may be non-negligible. I understand. On the other hand, as described in equation (2) in the explanation of the prior art example, howling suppression by the echo signal estimator has a long sampling period, and the cost can be reduced as the high frequency limit is lowered.

Therefore, in the present invention, howling suppression by the echo signal estimator is performed as shown in characteristic 33 in FIG.
The band is below 1.5KHz and occurs in a specific sound field.
Howling from 1.5 KHz to 3 KHz is suppressed by increasing the cutoff frequency of the high-pass filter from characteristic 34 to characteristic 35, as shown in FIG. In the case of a sound field, even if frequency components in the narrow midrange band are missing, it does not feel unnatural.

As described above, according to this embodiment, a howling suppression device with high howling suppression effect can be realized at low cost.

FIG. 5 shows a block diagram of a howling suppression device according to a second embodiment of the present invention. In FIG. 5, 51 is a howling detection section, 52
When the howling detection unit 51 detects howling, the high pass filter (HPF) 22
This is a cutoff frequency adjusting section for increasing the cutoff frequency of the second embodiment, and the other configurations are exactly the same as those of the first embodiment.

The howling suppression device of this embodiment configured as described above automatically adjusts the cutoff frequency, which was performed in the first embodiment.
Various methods can be considered to detect howling, but even if it is determined that howling is occurring due to the detection of sound pressure above a certain level when there is no input, the present invention can sufficiently achieve its purpose. can.

FIG. 6 shows a block diagram of a howling suppression device according to a third embodiment of the present invention. In FIG. 6, 61 is a second howling detection section;
62 is a damping section that increases the amount of attenuation when howling is detected by the howling detection section 61;
The other configurations are exactly the same as those of the second embodiment. The howling suppression device of this embodiment configured as described above makes it possible to automatically adjust the gain of the loudspeaker system focusing on howling in the low frequency range, which was impossible in the second embodiment, and is extremely effective. It is easy to operate and has great effects.

Effects of the Invention The present invention suppresses howling in the low frequency range by canceling the echo using the echo signal estimator, and in the midrange by adjusting the cutoff frequency of the high-pass filter to form a frequency band that does not feel unnatural. A howling suppression device with high suppression effect can be realized at a low cost, and furthermore, an automatic adjustment function combined with howling detection can realize a howling suppression device with high operability.

[Brief explanation of the drawing]

Fig. 1 is a block diagram of a conventional howling suppression device, Fig. 2 is a block diagram of a howling suppression device according to the first embodiment of the present invention, and Fig. 3 is a block diagram of a system between a loudspeaker and a microphone measured in an auditorium. A frequency transfer characteristic diagram, FIG. 4 is a frequency characteristic diagram of each part of the howling suppression device in the first embodiment of the present invention, FIG. 5 is a block diagram of the howling suppression device in the second embodiment of the present invention, and FIG. This figure is a block diagram of the howling suppressing device in the embodiment of FIG. 3 of the present invention. 3...Subtraction section, 21...Low pass filter, 2
2... High pass filter, 23... Cutoff frequency adjustment section, 24... Transfer characteristic estimation section, 26...
Echo signal estimator, 27... adder.

Claims (1)

[Scope of Claims] 1. A low-pass filter that extracts low-frequency components of the output of a microphone, a high-pass filter having a cutoff frequency higher than a cutoff frequency of the low-pass filter that extracts high-frequency components of the output of the microphone, and the high-pass filter that extracts the high-frequency components of the output of the microphone. a cutoff frequency adjustment section that adjusts the cutoff frequency of the speaker; a transfer characteristic estimation section that estimates the transfer characteristic of a low frequency range that is a pass band of the low-pass filter of the reverberation signal that circulates from the speaker to the microphone; and an input signal to the speaker; a reverberant signal estimator that approximately synthesizes a low-frequency reverberant signal circulating from the speaker to the microphone based on the estimated transfer characteristic; a subtractor that subtracts the output of the reverberant signal estimator from the output of the low-pass filter; A howling suppression device comprising: an addition section that adds the output of the subtraction section and the output of the high-pass filter. 2. Claim 1, characterized in that a howling detection unit is provided to detect howling in the passband of the high-pass filter, and when howling is detected by the howling detection unit, the cut-off frequency of the high-pass filter is increased. The howling suppression device described. 3. An attenuation section that attenuates the output of the addition section and a second howling detection section that detects howling in the passband of the low-pass filter are provided, and when howling is detected by the howling detection section, the attenuation of the attenuation section is 2. The howling suppression device according to claim 1, wherein the amount of howling is increased.