JPH09116361A - Automatic volume control equipment - Google Patents

Abstract

PROBLEM TO BE SOLVED: To enable loud speaking with high degree of clearness corresponding to a noise level even on the condition that the noise level fluctuates. SOLUTION: An audio signal is inputted from a loud speaking signal input means 1, an amplification level is controlled by an autovolume 2, and a loud speaking sound is outputted through an amplifier 3 and a speaker 4. The noise at a reproducing place is detected by a noise detecting microphone 5, and the band of the detected noise is divided by a band dividing filter 6. Next, the optimum hearing level of the loud speaking sound is predicted while using an optimum hearing level predicting means 7A composed of a neural network. Corresponding to the predicted level, a volume level control means 8 controls the level of the autovolume 2. Thus, the loud speaking sound can be heard at the optimum hearing level.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an automatic volume control device for reproducing sound with a high degree of clarity even in a noisy state when expanding a voice.

[0002]

2. Description of the Related Art Conventionally, for the purpose of reproducing a voice with high clarity even under a noisy condition, an automatic sound volume for detecting a noise level at a reproducing place and automatically controlling a volume of a loud sound according to its loudness. A control device has been proposed.

FIG. 8 is a block diagram showing a configuration example of such a conventional automatic volume control device. In FIG. 8, a loud sound signal input means 1 is a means for inputting a voice signal to be loud. The auto volume 2 is a circuit that adjusts the amplification level of the audio signal. Amplifier 3 is auto volume 2
It is an amplifier that amplifies the signal from. The speaker 4 reproduces the signal amplified by the amplifier 3.

The noise detection microphone 5 is a microphone for detecting the noise level. The noise level detection means 10A is a circuit that calculates the noise level based on the signal from the noise detection microphone 5. The volume level control means 8 is a means for controlling the amplification level of the automatic volume control 2 according to the calculation result of the noise level detection means 10A.

In the conventional automatic volume control device thus constructed, the audio signal input from the loudspeaker signal input means 1 is adjusted to an appropriate size by the auto volume 2.
It is reproduced from the speaker 4. The noise detection microphone 5 picks up a sound including noise and gives it to the noise level detection means 10A. Then, the noise level detection means 10A calculates the magnitude of the noise level from the detected sound.

As a noise level detecting method in this case, if a noise level is detected under a condition with a loud sound, the loudness of only the noise cannot be calculated. Therefore, it is determined whether or not there is a loud sound, and there is no loud sound. Calculate the noise level at the part.
The calculated noise level is input to the volume level control means 8. The volume level control means 8 sets the amplification level of the automatic volume 2 (hereinafter, simply referred to as a level) so as to amplify the audio signal with a constant level difference from the noise level regardless of the type of noise. In this way, the level of the automatic volume control 2 is changed according to the noise level, so that the voice signal can be amplified with high clarity even in a noisy condition.

[0007]

However, in the above-described conventional configuration, the level of the automatic volume 2 is the same as the noise level regardless of the type of noise, such as the cheering of the audience or the applause. Is set. However, even if the noise level is the same, if the type of noise is different, there will be a considerable difference in actual hearing, and there will be a problem that the loud sound is too loud and noisy, or conversely, it is too quiet and the intelligibility drops. It was

The present invention has been made in view of the above-mentioned conventional problems, and it is
It is an object of the present invention to provide an automatic volume control device capable of reproducing at an optimum sound level.

[0009]

According to the invention of claim 1 of the present application, a loudspeaker signal input means for inputting a voice signal to be amplified, and an automatic control for controlling an amplification level of the voice signal inputted from the loudspeaker signal input means. Based on a volume, a noise detection microphone that detects noise in a sound reproduction location, a band division filter that divides the noise detected by the noise detection microphone into bands, and a signal from the band division filter, a neural network is used. An optimal listening level predicting means for predicting an optimal listening level according to the above, and a volume level control means for controlling an amplification level of the auto volume with the optimal listening level predicted by the optimal listening level predicting means as an input. It is characterized by.

According to a second aspect of the present invention, a voice signal input means for inputting a voice signal to be voiced, an automatic volume control for controlling an amplification level of the voice signal input from the voice signal input means, and a voice reproduction. A noise detection microphone that detects noise in a location, a band division filter that divides the noise detected by the noise detection microphone into bands, and a multiple regression based on the signal from the band division filter and the signal from the noise detection microphone Optimum listening level predicting means for predicting the optimal listening level using analysis, based on the optimal listening level predicted by the optimal listening level predicting means, volume level control means for controlling the amplification level of the automatic volume,
It is characterized by including.

According to a third aspect of the present invention, a loudspeaker signal input means for inputting a voice signal to be amplified, an auto volume for controlling an amplification level of the voice signal input from the loudspeaker signal input means, and a voice reproduction are provided. A noise detection microphone that detects noise at a place, a band division filter that divides the noise detected by the noise detection microphone into bands, and a signal from the band division filter is used to predict an optimal listening level using a neural network. An optimal listening level predicting means for determining whether or not the optimal listening level predicted by the optimal listening level predicting means is appropriate, and outputs a correction value if the predicted values greatly differ, and the predicted value Volume level control means for controlling the amplification level of the auto volume based on the correction value from the correction means. Than it is.

According to a fourth aspect of the present invention, a loudspeaker signal input means for inputting a voice signal to be amplified, an auto volume for controlling an amplification level of the voice signal inputted from the loudspeaker signal input means, and the auto volume are provided. A speaker that outputs the audio signal output by the speaker as a sound, a noise detection microphone that detects noise in a sound reproduction place where the speaker is installed, and a first band that divides the noise detected by the noise detection microphone into bands. A division filter, a delay device for delaying an output signal of the loudspeaker signal input means by a propagation time from the speaker to the noise detecting microphone, and a delay filter having the same configuration as the first band division filter, Second band-division filter for band-dividing the output of the converter, and the first band-division filter from the output of the second band-division filter for each band A noise level detecting means for calculating only the noise level by subtracting a value obtained by multiplying the output value by a loudspeaking coefficient, and a neural network is used to predict the optimum listening level based on the signal from the noise level detecting means. It is characterized by comprising an optimum listening level predicting means and a volume level controlling means for controlling the amplification level of the automatic volume based on the output from the optimum listening level predicting means.

According to a fifth aspect of the present invention, a voice signal input means for inputting a voice signal to be amplified, an auto equalizer for controlling the frequency characteristic of the voice signal input from the voice signal input means, and a voice reproduction are provided. A noise detection microphone that detects noise in a place, a band division filter that divides the noise detected by the noise detection microphone into bands, and an optimal listening level and frequency using a neural network based on the signal from the band division filter. Optimal listening level predicting means for predicting characteristics, based on the optimal listening characteristics and frequency characteristics predicted by the optimum listening level predicting means, frequency characteristic control means for controlling the frequency characteristics of the auto equalizer,
It is characterized by including.

As described above, the automatic volume control device according to claims 1 and 2 of the present application inputs the voice signal from the loud sound input means, detects the noise level from the noise detection microphone installed at the reproduction place, and detects the noise level. Is band-divided by a band-division filter. The optimum listening level prediction means uses the band-divided signal to predict the optimum listening level according to the noise level. The amplification level of the auto volume is controlled according to the predicted value, and the sound is amplified at the optimum listening level according to the noise level.

The automatic volume control device according to claim 3 of the present application judges whether the optimum listening level predicted by the optimum listening level predicting means is a proper value, and if there is a large difference, the predicted value correcting means corrects the listening level. . This makes it possible to eliminate the change in the loud sound due to the error in the predicted value.

In the automatic volume control device according to claim 4 of the present application, when the loud sound and the noise level overlap, the noise level detection means calculates only the noise level. In this way, the volume level is prevented from increasing due to the influence of the loud sound.

In the automatic volume control device according to claim 5 of the present application, the optimum listening characteristic predicting means predicts the optimum listening characteristic and the optimum listening level, and the frequency characteristic controlling means controls the frequency characteristic of the loud sound. In this way, even if the noise level changes, a loud sound with good intelligibility is output.

[0018]

DESCRIPTION OF THE PREFERRED EMBODIMENTS An automatic volume control device according to a first embodiment (claim 1) of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram showing the configuration of the automatic volume control device according to the present embodiment. Parts having the same functions as those of the conventional one are designated by the same reference numerals, and their description will be omitted. The band division filter 6 is a filter that divides a signal due to noise detected by the noise detection microphone 5 into bands of, for example, 500 Hz, 1 KHz, and 2 KHz. Optimal listening level prediction means 7
A is a means for predicting the optimum listening level using a neural network based on the signal from the band division filter 6.
The volume level control means 8 is means for controlling the level of the automatic volume 2 according to the prediction result of the optimum listening level prediction means 7A.

The operation of the thus-configured automatic volume control device will be described. When a voice signal to be amplified is input from the voice amplification signal input means 1 in FIG. 1, the signal is controlled to an optimum level by the auto volume 2, and the amplifier 3
Is amplified and reproduced from the speaker 4. On the other hand, the noise signal detected by the noise detection microphone 5 is divided into a plurality of bands by the band division filter 6 and input to the optimum listening level prediction means 7A.

The optimum listening level predicting means 7A is composed of a neural network as shown in FIG. The neural network shown in FIG. 2 is formed by connecting a plurality of units, and has a three-layer structure here. The neural network of this embodiment has three input layers A, four intermediate layers B, and one output layer C. The input layer A inputs the respective components of 500 Hz, 1 KHz and 2 KHz divided by the band division filter 6. One unit has multiple inputs and one output, and the value x taken by each unit is represented as a sum (weight sum) X of weighted input values, and the unit output value here is f (X). It is expressed by the function.

The function f (X) used in this embodiment is a sigmoid function represented by the following equation (1).

(Equation 1) Further, the neural net used this time is of a hierarchical type, and the signal from the band division filter 6 is input to the input layer A, and the predicted value of the optimum listening level is output.

In the learning method for obtaining the coupling coefficient between the layers of the neural network, the actual noise is first input to the optimum listening level predicting means 7A to obtain the output value. Next, from the difference between the obtained signal and the correct output signal obtained in advance,
The weights connecting each unit are recalculated and corrected. Learning is advanced while repeating this processing, and the coupling coefficient of the neural network is obtained. A method using a Kalman filter was used as an actual learning method. As for the optimum listening level, a listening experiment is conducted in advance to find the relationship between the optimum listening levels according to the noise levels of several kinds of noises. The coupling coefficient obtained by such a method is set between the layers of the optimum listening level prediction means 7A.

The optimum listening level predicting means 7A predicts the optimum listening level by using the above neural network. The predicted level is given to the volume level control means 8, and the volume level control means 8 controls the level of the automatic volume 2. As described above, the loud sound is reproduced at the level set in the auto volume 2. This loud sound can be heard with high clarity even under noisy conditions. It should be noted that although a three-layer structure of a new Brunet with three inputs is used here, different numbers of inputs and layers may be used.

Next, an automatic volume control device according to a second embodiment (claim 2) of the present invention will be described with reference to the drawings. FIG. 3 is a block diagram showing the configuration of the automatic volume control device according to the present embodiment. The parts having the same functions as those in the first embodiment are designated by the same reference numerals and their description is omitted. First
Unlike the embodiment, the optimum listening level predicting means 7B shown in FIG.
Is configured such that the signal from the noise detection microphone 5 is directly input together with the signal from the band division filter 6.

The operation of the automatic volume control device having such a configuration will be described. The noise detected by the noise detection microphone 5 is input to the band division filter 6, and 315 Hz to 6.3 KH
The z band is divided into subbands divided by 1/3 octave. Then, this signal is input to the optimum listening level prediction means 7B. The signal from the noise detection microphone 5 is also directly input to the optimum listening level prediction means 7B.

FIG. 4 is a diagram showing the structure of the optimum listening level predicting means 7B. The optimum listening level predicting means 7B is a signal for each band from the band division filter 6 and the noise detection microphone 5
The signal itself is used to perform the multiple regression calculation shown in the equation (2) to predict the optimum listening level.

(Equation 2) Sopt: Optimal listening level N315-N6K: Noise level a0, a315-a6.3K: Regression coefficient of each band calculated by multiple regression analysis The level predicted by the optimal listening level predicting means 7B is the same as in the first embodiment. The automatic volume 2 is set by the method of.

In this way, the optimum listening level predicting means 7B
However, by predicting the optimum listening level using the multiple regression equation, it is possible to reproduce a sound with high clarity without being masked by this noise even under a noisy condition. Note that, although the multiple regression calculation is performed here using 14 bands in each 1/3 octave, different bandwidths and number of bands may be used.

Next, an automatic volume control device according to a third embodiment (claim 3) of the present invention will be described with reference to the drawings. FIG. 5 is a block diagram showing the configuration of the automatic volume control device according to the present embodiment. Parts having the same functions as those of the first embodiment are designated by the same reference numerals and their description is omitted. The automatic volume control device of the present embodiment is characterized in that a prediction value correction means 9 is provided between the optimum listening level prediction means 7A and the volume level control means 8 in addition to the components shown in FIG.

The operation of the thus-configured automatic volume control device will be described. The method of predicting the optimum listening level is the same as in the first embodiment. The value predicted by the optimum listening level prediction means 7A is input to the predicted value correction means 9. The predicted value correction means 9 judges whether or not the predicted value is a value greatly different from the level obtained immediately before, and if it is significantly different, corrects the value to a value close to the previous level and corrects it by the volume level control means 8. Give a value. Then, the volume level control means 8
Controls the level of the auto volume 2 so that the volume changes naturally as in the first embodiment.

With respect to the point that the prediction value error may become large when a neural network is used, by correcting the prediction value error, a sound that is easy to hear is reproduced without greatly deviating from the actual optimum listening level. We were able to.

Next, an automatic volume control device according to a fourth embodiment (claim 4) of the present invention will be described with reference to the drawings. FIG. 6 is a block diagram showing the configuration of the automatic volume control device of the present embodiment. The parts having the same functions as those of the first and third embodiments are designated by the same reference numerals and their description is omitted. The automatic volume control device according to the present embodiment, in addition to the components shown in FIG. 5, includes a noise level detecting means 10B and a delay device 11.
Respectively, two sets of band division filters 6A,
6B is provided. The delay device 11 delays the signal from the loudspeaker signal input means 1, and its output is given to the first band division filter 6A. The output of the noise detection microphone 5 is given to the second band division filter 6B.

The operation of the automatic volume control device having such a configuration will be described. The audio signal input from the loudspeaker signal input means 1 is controlled to an appropriate level by the auto volume 2,
A loud voice is given through the speaker 4. On the other hand, the input audio signal is delayed by the delay device 11. This delay time is the time during which sound propagates from the speaker 4 to the noise detection microphone 5. The delayed audio signal is band-divided by the band-division filter 6A and input to the noise level detecting means 10B.

Next, the signal picked up by the noise detection microphone 5 is divided into several bands by the second band division filter 6B and input to the noise level detection means 10B. The band divided here is the same as the band divided by the first band division filter 6A. Noise level detection means 10B
Calculates the actual noise level for each band using the signals of the band division filters 6A and 6B. As a calculation method, the relationship between the level L of the auto volume 2 and the loudness M of the loud sound entering the noise detection microphone 5 is obtained in advance as a loudness coefficient α. The method of obtaining the loudness coefficient α is shown in Expression (3).

(Equation 3) α: Loudness coefficient L: Auto volume level M: Noise level detection microphone level

Next, the noise level detecting means 10B subtracts the value obtained by multiplying the value of the band dividing filter 6A by the loudness coefficient α for each band. By this signal processing, the loud sound is removed from the signal in the state where the noise and the loud sound are mixed, and only the noise signal is extracted. Hereinafter, using the signal for each band obtained by the noise level detection means 10B,
The level of the automatic volume control 2 is controlled by the same operation as in the first embodiment.

As described above, even when the loud sound is mixed with the noise, only the noise level is obtained, and the automatic volume 2 is controlled using only the noise level, so that the optimum volume level is not affected by the loud sound. You can make a loud sound with. The gain of the amplifier 3 is constant here.

Next, an automatic volume control device according to a fifth embodiment (claim 5) of the present invention will be described with reference to the drawings. FIG. 7 is a block diagram showing the configuration of the automatic volume control device according to the present embodiment. Parts having the same functions as those of the first embodiment are designated by the same reference numerals and their description will be omitted. The auto equalizer 14 is a circuit that corrects the frequency characteristics of the audio signal output from the loudspeaker signal input means 1, and its output is given to the amplifier 3. Optimal listening characteristic prediction means 15
When the signal of the band division filter 6 is input, the neural network is used to obtain the optimum listening characteristic according to the frequency characteristic of noise. The frequency characteristic control means 16 is means for giving a control signal of the frequency characteristic to the auto equalizer 14 based on the predicted value of the optimum frequency characteristic predicting means 15.

The operation of the automatic volume control device thus configured will be described. When the voice signal to be amplified is input from the voice amplification signal input means 1 of FIG. 7, the signal is controlled by the auto equalizer 14 to have an optimum frequency characteristic. This signal is amplified by the amplifier 3 and reproduced from the speaker 4. On the other hand, the noise signal detected by the noise detection microphone 5 is divided into a plurality of bands by the band division filter 6 and input to the optimum listening characteristic prediction means 15. As the optimum listening characteristic, it is assumed that the noise is classified into some patterns and the characteristics obtained from psychological experiments are used for each pattern. The optimum frequency characteristic obtained here is set in the auto equalizer 14 via the frequency characteristic control means 16.

The neural network used in this embodiment is basically the same as the neural network used in the first embodiment. As described above, by obtaining the optimum frequency characteristic according to the noise by using the optimum frequency characteristic predicting means 15, it is possible to perform a clear loud sound even in the noise.

[0039]

As described above, the automatic volume control device of the present invention detects noise at the reproduction location of the loud sound and divides the detected noise into frequency band components. Then, each band component is input to the optimum listening level predicting means or the optimum listening characteristic predicting means to predict the optimum listening level according to the noise level. By using the level predicted here, the volume of the loud sound can be varied so as to give a natural loud voice, so that the volume level can be controlled in real time. That is, the loud sound can be heard with high clarity according to the noise level and type.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of an automatic volume control device according to a first embodiment of the present invention.

FIG. 2 is a configuration diagram of a neural network used in the optimum listening level prediction means of the first embodiment.

FIG. 3 is a block diagram showing a configuration of an automatic volume control device according to a second embodiment of the present invention.

FIG. 4 is an explanatory diagram showing a calculation process of a multiple regression equation used in the optimum listening level prediction means of the second embodiment.

FIG. 5 is a block diagram showing a configuration of an automatic volume control device according to a third embodiment of the present invention.

FIG. 6 is a block diagram showing a configuration of an automatic volume control device according to a fourth embodiment of the present invention.

FIG. 7 is a block diagram showing a configuration of an automatic volume control device according to a fifth embodiment of the present invention.

FIG. 8 is a block diagram showing a configuration example of a conventional automatic volume control device.

[Explanation of symbols]

 1 loudspeaker signal input means 2 auto volume 3 amplifier 4 speaker 5 noise detection microphone 6, 6A, 6B band division filter 7, 7A, 7B optimal listening level prediction means 8 volume level control means 9 predicted value correction means 10A, 10B noise level detection Means 11 Delay device 14 Auto equalizer 15 Optimal listening characteristic predicting means 16 Frequency characteristic controlling means

Claims (5)

[Claims] 1. A loudspeaker signal inputting means for inputting a speech signal to be loudened, an automatic volume control for controlling an amplification level of the speech signal inputted from the loudspeaking signal input means, and a noise for detecting noise at a sound reproducing place. A detection microphone, a band division filter that divides the noise detected by the noise detection microphone into bands, and an optimal listening level prediction unit that predicts an optimal listening level using a neural network based on the signal from the band division filter, An automatic volume control device comprising: a volume level control unit that controls the amplification level of the auto volume by using the optimum listening level predicted by the optimum listening level prediction unit as an input. 2. A loudspeaker signal input means for inputting a loudspeaker signal to be amplified, an automatic volume control for controlling an amplification level of the sound signal inputted from the loudspeaker signal input means, and a noise for detecting noise at a sound reproduction place. A detection microphone, a band division filter that divides the noise detected by the noise detection microphone into bands, and an optimal listening level using multiple regression analysis based on the signal from the band division filter and the signal from the noise detection microphone. And an audio volume level control means for controlling the amplification level of the automatic volume based on the optimal listening level predicted by the optimal listening level predicting means. Automatic volume control device. 3. A loudspeaker signal input means for inputting a loudspeaker signal to be amplified, an automatic volume control for controlling an amplification level of the sound signal inputted from the loudspeaker signal input means, and a noise for detecting noise at a sound reproduction place. A detection microphone, a band division filter that divides the noise detected by the noise detection microphone into bands, and an optimal listening level prediction unit that predicts an optimal listening level using a neural network based on the signal from the band division filter, , A prediction value correction unit that determines whether the optimum listening level predicted by the optimum listening level prediction unit is appropriate, and outputs a correction value when the prediction value is significantly different, and a correction value from the prediction value correction unit. And a volume level control means for controlling an amplification level of the auto volume, the automatic volume control device. 4. A voice amplification signal input means for inputting a voice signal to be amplified, an auto volume control for controlling an amplification level of the voice signal input from the voice amplification signal input means, and a voice signal output by the auto volume. A noise detection microphone that detects noise in a sound reproduction place where the speaker is installed, a first band division filter that band-divides the noise detected by the noise detection microphone, and the loud signal A delay device that delays the output signal of the input means by a propagation time from the speaker to the noise detection microphone, and the same configuration as the first band division filter, and divides the output of the delay device into bands. A second band division filter, and an output coefficient of each band from the output of the second band division filter to the output value of the first band division filter A noise level detecting means for calculating only a noise level by subtracting the multiplied value, and an optimum listening level predicting means for predicting an optimum listening level using a neural network based on a signal from the noise level detecting means. And an audio volume control means for controlling the amplification level of the auto volume based on the output from the optimum listening level prediction means. 5. A loudspeaker signal input means for inputting a speech signal to be loudened, an auto equalizer for controlling the frequency characteristic of the speech signal inputted from the loudspeaker signal input means, and a noise for detecting noise at a voice reproduction place. A detection microphone, a band division filter that divides the noise detected by the noise detection microphone into bands, and an optimal listening level that predicts an optimal listening level and frequency characteristics using a neural network based on the signal from the band division filter. An automatic volume comprising: level predicting means; and frequency characteristic control means for controlling the frequency characteristic of the auto equalizer based on the optimum listening characteristic and frequency characteristic predicted by the optimum listening level predicting means. Control device.