JPH0632538B2 - Howling detector - Google Patents

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a howling detection device in a loudspeaker system at a lecture or a conference, a teleconference system, or the like.

2. Description of the Related Art In recent years, in addition to lectures and conferences, systems using a loudspeaker such as a teleconference system have been developed, and a clear and natural voice transmission with a sufficient loudness is desired. .

However, in reality, since there is a risk of howling, it is not possible to sufficiently raise the loudness level, and only the loudness level is low and the degree of intelligibility is poor. Because howling grows rapidly once it begins to grow,
The maximum output of the loudspeaker is reached in a short time. Therefore, howling should never occur during operation of the loudspeaker. For this reason, the gain of the loudspeaker is usually set to be sufficiently low in advance. In particular, when used by general users such as in lectures and conferences, and when the acoustic characteristics of the sound field change drastically, the gain has to be suppressed even lower.

That is, even if an attempt is made to operate the gain of the loudspeaker with a maximum gain within a range that does not cause howling in order to obtain a sufficient loudness volume, howling occurs even if the acoustic characteristic of the sound field is slightly changed, Operation with this gain is not possible. However, in order to obtain a sufficient loudness volume, the loudspeaker must be operated with this maximum gain. In order to realize the operation in this state, it is necessary to detect howling at an early stage of its generation and automatically control the gain and frequency characteristics of the loudspeaker.

Research on this technique has been made in the past, and one example is the howling suppression device by Matsumoto et al. (Proceedings of the Spring Meeting of the Acoustical Society of Japan 1983, P313) The conventional howling detection apparatus described above will be described with reference to the drawings.

FIG. 5 shows the configuration of a conventional loudspeaker system incorporating a howling detection device. In FIG. 5, 51 is a microphone, 52 is a howling detection device, 521 is a frequency analysis unit for performing frequency analysis of the output signal of the microphone,
Reference numeral 522 is a determination unit that determines the occurrence of howling from the frequency analysis result, 53 is a gain adjusting unit that reduces the gain of the loudspeaker when howling occurs, 54 is a power amplifier, and 55 is a speaker. In the loudspeaker apparatus of FIG. 5, the howling detection apparatus 52 is composed of a frequency analysis unit 521 and a determination unit 522.

The operation of the loudspeaker incorporating the howling detection apparatus configured as described above will be described below.

The voice signal of the speaker is picked up by the microphone 51, amplified by the power amplifier 54, and then amplified by the speaker 55.
Since the acoustic coupling between the microphone 51 and the speaker 55 cannot be reduced to zero, the amplified voice signal is transmitted to the microphone 5 again.
After being picked up by 1 and amplified, it is again amplified by the speaker 55, and a field back loop is formed. This is the cause of howling, and causes an oscillating phenomenon, that is, howling at a frequency where the gain of the field back loop exceeds 1.

On the other hand, the output signal of the microphone 51 is the frequency analysis unit 52.
The frequency analysis is performed at 1, and the result is sent to the determination unit 522. The determination unit 522 uses the property that the howling sound is closer to a pure tone than the sound, and if a frequency analysis result that energy is concentrated in a narrow frequency band is obtained, it is determined that howling has occurred. The judgment result is sent to the gain adjusting section 53. This criterion is whether the ratio of the energy in a certain frequency band to the energy in the frequency bands on both sides of the frequency band is larger or smaller than a preset threshold value. If it is larger, howling occurs in that frequency band. It is determined that If it is determined that howling is occurring, then the gain adjusting unit 53 reduces the gain of the loudspeaker system to suppress howling.

Problems to be Solved by the Invention However, with the above-mentioned howling detection device, the howling cannot be detected until the howling has grown considerably.
This is because the voice is also quite pure in the vowel part, and it was impossible to completely separate and detect the voice and the howling sound in the early stage of howling growth. For this reason, the conventional loudspeaker incorporating the howling detection apparatus has a drawback that considerably howling occurs, though in a short time.

In view of the above problems, the present invention provides a howling detection device that detects howling when it is about to occur or at an early stage of its growth.

Means for Solving the Problems In order to achieve this object, a howling detection apparatus of the present invention is a section detection unit that detects a section immediately after a syllable included in a voice signal of a loudspeaker, and a section immediately after this syllable. If the attenuation characteristic calculation unit that calculates the attenuation characteristic of the signal and the attenuation characteristic calculated by the attenuation characteristic calculation unit are different from normal voice,
And a determination unit that determines that howling is occurring.

Function With this configuration, howling can be detected by using the attenuation characteristics of the signal in the section immediately after the syllable, that is, the attenuation time and the attenuation rate, which are most sensitive to howling. Is dramatically improved. Since howling can be detected at an early stage of its growth, which is still inconspicuous, by combining this detecting device with a howling suppressing device such as automatic control of the loudspeaker gain, a loudspeaker completely preventing howling can be realized. With this loudspeaker, howling can be detected and suppressed in advance, so that the initial setting value of the gain of the loudspeaker can be set near the howling limit, and the loudness can be increased.

Embodiment An embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 shows the structure of a howling detection apparatus according to the first embodiment of the present invention. In FIG. 1, 1 is an input terminal of a signal, 2 is an amplitude calculating section for obtaining a short-term average amplitude of the input signal, and 3 is a section detecting section for detecting a syllable section of a voice in the input signal from the short-time average amplitude. Numeral 4 is a syllable amplitude calculator 5 for obtaining the maximum value of the amplitude in the syllable section from the short-time average amplitude sent from the amplitude calculator 2 and the section detection result sent from the section detector 3. The section immediately after the syllable from the average amplitude sent from the calculation section 2, the section detection result sent from the section detection section 3, and the maximum amplitude value in the syllable section sent from the syllable amplitude calculation section 4. An attenuation characteristic calculation unit 6 for obtaining the attenuation time of the signal is a determination unit for determining that howling is about to occur when the attenuation time sent from the attenuation characteristic calculation unit 5 is longer than a predetermined threshold value. , 7 are output terminals for the determination result.

The operation of the howling detection apparatus configured as described above will be described below.

First, the output signal of the microphone of the loudspeaker is input terminal 1
Is supplied to the amplitude calculator 2 and the short-term average amplitude is obtained. The section detection unit 3 calculates the differential value of this short-time average amplitude, and determines the section from the time when the polarity of this differential value becomes positive to the time when it is negatively inverted as a syllable section. The syllable amplitude calculation unit 4 obtains the maximum value of the short-term average amplitude in the section determined by the section detection unit 3 as the syllable section, and sends the result to the attenuation characteristic calculation unit 5. The attenuation characteristic calculation unit 5 monitors the change in the short-time average amplitude of the signal after the syllable section detected by the section detection unit 3, and decays from the maximum value of the syllable signal amplitude calculated immediately before to a predetermined rate. That is, the decay time is measured. Finally, when the attenuation time sent from the attenuation characteristic calculation unit 5 is longer than that of normal voice, the determination unit 6 determines that howling is about to occur, and the output terminal 7 outputs the determination result.

Next, the principle of this howling detection device will be described in detail.

The signal waveform of the input terminal 1 is shown in FIG. In FIG. 3, 31 is an input signal waveform in a state where no howling is occurring, 32 is a howling margin of 3 dB,
The input signal waveform is in a slightly howling state. In the state of the howling margin of 3 dB, the howling did not continue or grow, and a weak howling sound was sometimes felt, and the howling sound also immediately attenuated.

Comparing the two waveforms, it can be seen that immediately before howling, a signal component due to howling is generated in the section immediately after the syllable of the speech, and exhibits an attenuation characteristic different from that of normal speech. Therefore, if the attenuation characteristic of the signal in the section immediately after this syllable is used as a howling determination element, it becomes possible to perform high-precision howling detection that can be detected before the howling. The present invention realizes exactly this principle.

Next, the role and operation of the section detector 3 will be described in more detail.

In order to obtain the attenuation characteristic, it is first necessary to detect the section immediately after the syllable of the voice in the input signal, that is, the attenuation section. Many experiments have found that this section detection is possible by monitoring the amplitude change of the signal of the loudspeaker.

As one of the methods, a method of utilizing a differential value of amplitude can be considered. For example, the start point of the syllable section is when the signal amplitude sharply increases, that is, when the differential value becomes positive, and the end point thereof is when the signal amplitude starts to sharply or monotonically decrease, that is, the differential value. There is a method in which the value becomes negative and the section immediately after the syllable is a section after this syllable section. Since the syllable section and the section immediately after the syllable are in contact with each other, the detection of the syllable section is also the detection of the section immediately after the syllable. In the actual voice, the amplitude sharply increases at the beginning of the syllable, and thereafter, the amplitude gradually decreases as a whole, though it slightly changes. Therefore, according to the above criteria, the start point can be accurately obtained, but the end point cannot be obtained accurately. However, in practice, the above determination method is not a big problem, and in the embodiment shown in FIG. 1, this method is used to detect a syllable section and a section immediately following the syllable section.

FIG. 4 shows the actual operation of this syllable section detection method, that is, the section detection method immediately after the syllable. In FIG. 4, 41 is the same as the input signal waveform 32 of the howling margin 3 dB in FIG. 3, and 42 is the input signal waveform 41 obtained by the amplitude calculator 2.
Is a short-time average amplitude, and 43 is a differential value of the short-time average amplitude 42 calculated in the section detection unit 3. This differential value 4
Comparing 3 and the input signal waveform 41, it can be seen that the differential value 43 is always positive in the syllable section, and the differential value 43 is always negative in the section immediately after the syllable. Therefore, it is possible to detect the section based on whether the differential value is positive or negative.

As described above, according to the present embodiment, the howling detection accuracy can be dramatically improved by providing the section detection unit, the attenuation characteristic calculation unit, and the determination unit. By combining with a howling suppression device such as automatic control of system gain, it is possible to realize a loudspeaker that completely prevents howling.

A second embodiment of the present invention will be described below with reference to the drawings.

FIG. 2 shows the configuration of a howling detection apparatus showing a second embodiment of the present invention. In FIG. 2, 1 is an input terminal of a signal, 8 is a frequency band division unit that divides the input signal into four frequency bands, 2a to 2d are amplitude calculation units that obtain short-time average amplitudes of the respective band signals, and 3a to Reference numeral 3d denotes a section detection unit that detects a syllable section of a voice in a band signal from the short-time average amplitude, and reference numerals 4a to 4d denote short-time average amplitudes and section detection units 3a to 3d sent from the amplitude calculation units 2a to 2d. The syllabic amplitude calculation unit that obtains the maximum value of the amplitude of each band signal in the syllable section from the section detection result sent from
a to 5d are average amplitudes sent from the amplitude calculators 2a to 2d, section detection results sent from the section detectors 3a to 3d, and syllable sections sent from the syllable amplitude calculators 4a to 4d. The attenuation characteristic calculators 6a to 6d, which determine the attenuation time in each frequency band of the signal in the section immediately after the syllable from the maximum value of the amplitude of, are determined by the attenuation time sent from the attenuation characteristic calculators 5a to 5d. When the time is longer than the threshold value, the determining unit determines that howling is about to occur, and 7a to 7d are output terminals for the determination result. In FIG. 2, reference numerals 2a to 2d denote the amplitude calculation unit 2 of FIG.
3d is the section detection unit 3 in FIG. 1, 4a to 4d are syllable amplitude calculation units 4 in FIG. 1, 5a to 5d are attenuation characteristic calculation units 5 in FIG. 1, and 6a to 6d are in FIG. Judgment unit 6 and 7a to 7
d is exactly the same as the output terminal 7 in FIG.

The operation of the howling detection apparatus configured as described above will be described below.

First, the output signal of the microphone of the loudspeaker is input terminal 1
And is supplied to the frequency band division unit 8 and divided into four frequency band signals. After that, howling is detected for each frequency band by the same method as in the first embodiment. The howling determination result in each frequency band is output to the outside from the output terminals 7a to 7d.

As described above, the frequency band division unit 8 and the plurality of sets of the first
By providing the howling detection means of the embodiment of the figure,
It is possible to know in which frequency band howling is occurring.

Advantageous Effects of Invention The present invention has a section detection unit that detects a section immediately after a syllable included in a voice signal of a loudspeaker that amplifies a voice, and an attenuation characteristic calculation unit that calculates an attenuation characteristic of a signal immediately after a syllable.
If the attenuation characteristic calculated by the attenuation characteristic calculator is different from that of normal voice, it is composed of a determination unit that determines howling is occurring, so that the section immediately after the syllable is most sensitive to howling. It becomes possible to detect howling by using the attenuation characteristics of the signal, that is, the attenuation time and the attenuation rate, and it is possible to realize a howling detection apparatus with dramatically improved accuracy of howling detection.

Further, by providing a frequency band division unit and a plurality of sets of howling detection means for each divided frequency band, it is possible to know in which frequency band howling is occurring.

The howling detection device of the present invention can detect howling at an early stage of its growth where the howling is still inconspicuous. Therefore, by combining this detection device with a howling suppression device such as automatic control of loudspeaker gain, howling is completely prevented. It is possible to realize a loudspeaker. With this loudspeaker, howling can be detected in advance and suppressed, so the initial setting value of the gain of the loudspeaker can be set near the howling limit, the loudness of the loudspeaker can be improved, and its practical effect is great. .

[Brief description of drawings]

FIG. 1 is a block diagram of a howling detection apparatus according to a first embodiment of the present invention, FIG. 2 is a block diagram of the same second embodiment, and FIG. 3 is a waveform diagram of an input signal of the howling detection apparatus.
FIG. 4 is a diagram for explaining the principle of section detection, and FIG. 5 is a block diagram showing a configuration of a conventional howling detection device and a loudspeaker using the howling detection device. 2, 2a to 2d ... Amplitude calculation section, 3, 3a to 3d ... Section detection section, 4, 4a to 4d ... Syllable amplitude calculation section, 5, 5
a to 5d ... Attenuation characteristic calculation unit, 6, 6a to 6d ... Judgment unit, 8 ... Frequency division unit.

Claims (2)

[Claims] 1. A section detecting section for detecting a section immediately after a syllable included in a voice signal of a loudspeaker for amplifying a voice, an attenuation characteristic calculating section for calculating an attenuation characteristic of a signal immediately after the syllable, and A howling detection apparatus comprising: a determination unit that determines that howling is occurring when the attenuation characteristic calculated by the attenuation characteristic calculation unit is different from normal voice. 2. A frequency dividing unit for dividing a voice signal of a loudspeaker into a plurality of frequency bands, and howling is detected for each frequency band.
The howling detection apparatus according to the item.