JPH08250944A - Automatic sound volume control method and device executing this method - Google Patents

Abstract

PURPOSE: To obtain an automatic sound volume control method in which the sound volume of an object signal only is amplified and the sound volume of the object signal only is automatically controlled and to obtain the device executing this method. CONSTITUTION: The device is provided with a buffer circuit 201 segmenting an input signal for each short time block and storing it, and with a sound signal identification circuit 202 identifying a kind of an input signal to be an object signal from other signals, and also with a sound volume amplifier circuit 203 amplifying and controlling the segmented signal depending on the identification result of the sound signal identification circuit 202 and with a waveform superimposing circuit 204 multiplying a weight window of the block length with a signal of the short time block signal subject to sound volume amplification control.

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 method and an apparatus for implementing this method, and more particularly to an automatic volume control method for controlling the volume of only a target signal while minimizing the volume control distortion. An apparatus for performing the method.

[0002]

2. Description of the Related Art A conventional automatic gain control AGC for adaptively controlling a signal level or power of an input signal according to its magnitude amplifies a signal of a small level, and an appropriate signal of an extremely large level is suitable. The main purpose is to limit the level to No matter what kind of signal the input signal is, the volume control is performed using only the signal level or power information. This will be described with reference to FIG. 4. The input signal L _in (t) at time t applied to the signal input terminal 301 is controlled by the maximum value limiting circuit 303 after the signal level is controlled by the amplifier circuit 302. Amplitude limited,
The output signal L _out (t) is output from the signal output terminal 304. This output signal L _out (t) is supplied to the integrating circuit 30.
5 and the gain control characteristic setting circuit 306 are set as in Expression (1).

L _out (t) = F (y) * L _in (t) (1) Here, F (y) is given as a non-linear characteristic, and normally it is a characteristic that increases when the input signal is small. Is set as. As described above, the conventional AGC controls an extremely simple function of amplifying the input level to some extent only when the input level is low, based on the signal magnitude, that is, only the level information. Therefore, the AGC apparatus constructed by using such a conventional technique has a result of amplifying signals other than the target signal to be amplified, for example, residual noise or environmental noise in the voice coding system. Thus, when the conventional AGC device is applied to the input signal containing the residual noise or the environmental noise, the output signal becomes very difficult to hear.

[0004]

In the conventional example of the automatic volume control method as described above, attention is paid only to the magnitude of the input signal, that is, the signal level or signal power information, and the volume is automatically controlled by the magnitude. However, when applied to an input signal on which environmental noise and other noises are superimposed, it is very difficult to hear because it greatly amplifies the signal portion of the environmental noise, which results in the conventional automatic volume control device. Is considered to be one of the causes of deterioration of the quality of the output signal.

According to the present invention, the input voice signal is divided into short-interval frames each having a frame period T _{f of} about 50 ms, and the input signal of this frame unit is a target signal to be amplified or other signals. The present invention provides an automatic volume control method and a device for carrying out this method, in which the sound is discriminated, and the result is used to amplify the volume of only the target signal to automatically control the volume of only the target signal.

[0006]

For an input signal in which a target signal and other signals are mixed, the input signal is cut out for each short time interval, the type of the sequential signal is identified, and the weighting coefficient is applied to the interval. An automatic volume control method for controlling the volume was constructed. Then, when the cut out signal section is identified as an audio signal, an automatic volume control method for amplifying it is configured.

Further, an automatic volume control method is provided which attenuates the cut out signal section when it is identified as a noise signal. Furthermore, when the cut out signal section is identified as a voice signal, it is amplified and
An automatic volume control method was constructed to attenuate the cut out signal section when it was identified as a noise signal.

The weighting factor constitutes an automatic volume control method which is a weighting factor for the cosine window. Further, a buffer circuit 201 for extracting and accumulating an input signal for each short time period is provided, and an audio signal identification circuit 202 for identifying the type of the input signal as a target signal and a signal other than that is provided, and a cut-out signal is provided. A sound volume amplification circuit 203 that performs amplification control according to the identification result of the audio signal identification circuit 202 is provided, and a waveform superposition circuit 204 that multiplies a signal of a short time period for which sound volume amplification control is performed by a weighting window of the section length is provided. An automatic volume control device was constructed.

[0009]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT Referring to FIGS.
Explain. First, the outline of the present invention will be explained with reference to FIG.
Reveal Input signal that changes with time shown in Fig. (A)
Signal power and this as shown in Figure (b).
Period T _fWith a window length WL twice as long as
For each outgoing input signal, it is an audio signal
And whether it is a signal other than that,
The volume is increased only when it is determined that the signal is an audio signal. here
In addition, these are also cut out to input the short section whose volume is controlled.
Each of the force signals is multiplied by the cosine function weight,
It shifts only WL / 2 time and overlaps with the cut-out waveform of the previous time.
Can be matched. Weighting coefficient W of the cosine window at this time
(I) is defined by equation (2).

W (i) = A * [0.5−0.5 * cos (2πi / WL)] (2) Here, A is an amplification coefficient, and in the example of FIG. 1, the amplification coefficient A is 3. It is 0. Further, the window shape of the weighting factor is determined by i, i = 0.0 to WL, and the overlapping section has a window shape of 0.0 to 1.0 in the front frame and 1.0 to 0 in the rear frame. Since the sum of the two is always 1.0 as a result, the input signal waveform will not be distorted at all unless the volume is controlled. That is, as shown in FIG. 2, for example, the window shape W _d (t) of the weighting factor in frame (t) is 1.
The window shape W _d (t + 1) of the weighting factor in the next frame (t + 1) is 0.0 to 1.0. Therefore, when A = 1.0, the input audio signal is always multiplied by 1.0 as a weighting coefficient, so that the waveform remains unchanged. For the same reason, by using such a cosine window, even if the volume is controlled in the overlapping portion, it gradually changes from 1.0 to 0.0 and from 0.0 to 1.0. Since the connection changes, by appropriately selecting the length of WL, it becomes possible to suppress a large connection distortion as a result. That is, if the shape of the clipping window is a rectangle that is always 1, here, if the amplification coefficient A is other than 1.0, the shape of the weighting coefficient superimposed for each frame is temporally "stepwise". ",
The output waveform undergoes rapid deformation, resulting in a large deterioration of the output signal.

FIG. 3 shows an embodiment of the automatic volume control device of the present invention. In the volume control process by the automatic volume control device, first, the input signal is cut out by the cutout window length WL and is temporarily stored in the buffer 201. Next, the audio signal identification circuit 202 for the cut-out WL input signal
In step S1, voice discrimination is made as to whether it is the target signal to be amplified or another signal. Then, the amplification factor A _m volume amplifier 203 normally being set to 1.0, from 1 the clipped signal interval which is set in advance so when identified as a voice signal larger Gain A _s
This is applied to the volume amplification circuit 203. here,
When cut out signal section is identified as the noise signal may be a control which allowed to decrease the amplification factor A _m volume amplifier 203 to the amplification factor A _s made smaller than 1. Alternatively, the cut-out signal interval and the amplification factor A _s is made larger than 1 the amplification factor A _s when identified as a speech signal, the cut-out signal interval is identified to be a noise signal time can be the amplification factor a _m and the amplification factor a _s made smaller than 1. Finally, in the waveform superimposing circuit 204, the previous cosine coefficient is multiplied, and the waveform is superposed on the previous frame to obtain an output signal subjected to the automatic volume control processing. The power of this output signal is shown in FIG.

By the above automatic volume control processing, the volume is amplified only when the audio signal is input, and it is possible to realize an efficient volume control device which does not amplify unnecessary noise. The processing quality of the sound obtained by this volume control device depends largely on the sound identification rate, but according to the preliminary examination result, 90% or more of people consider that there is no problem in the processing quality. If approximately 90%
It is known that the above voice recognition rate is necessary. At present, the voice identification algorithm used in the voice identification circuit is HMM (Hidden Markov Model), VQ (Vector Quantization), NW (Neural Network) or a method of utilizing the analysis result of the harmonic structure of the signal and various other methods. Has been proposed, and a discrimination rate of about 90% or more can be usually obtained. Therefore, an appropriate method is selected from these voice identification methods in consideration of the scale of the volume control device and the desired processing quality.

As described above, according to the present invention, the performance of the automatic sound volume control device can be remarkably improved by identifying the input signal for each short time section and further multiplying the section by the weighting coefficient.

[0014]

As described above, according to the present invention, the input signal is discriminated on a frame-by-frame basis, the volume of only the target signal is controlled, and the overlapping section is always 1.0.
It is possible to control the volume of the target signal buried in the signals other than the target signal by suppressing deterioration of the sound quality as much as possible by performing continuous processing by connecting with a window function such that

[Brief description of drawings]

FIG. 1 is a diagram for explaining the outline of the present invention.

FIG. 2 is a diagram illustrating weighting coefficients.

FIG. 3 is a diagram for explaining an embodiment of the present invention.

FIG. 4 is a diagram illustrating a conventional example of an automatic gain control method.

[Explanation of symbols]

 201 buffer circuit 202 audio signal identification circuit 203 volume amplification circuit 204 waveform superposition circuit

Claims (6)

[Claims] 1. A sound volume is controlled by cutting out an input signal in which a target signal and other signals are mixed, for each short time interval, identifying the type of the successive signal, and further multiplying the interval by a weighting coefficient. An automatic volume control method characterized by the above. 2. The automatic volume control method according to claim 1, wherein, when the cut out signal section is identified as a voice signal, it is amplified. 3. The automatic volume control method according to claim 1, wherein when the cut out signal section is identified as a noise signal, the noise is attenuated. 4. The automatic volume control method according to claim 1, wherein when the cut out signal section is identified as a voice signal, the cut out signal section is amplified, and the cut out signal section is a noise signal. When it is identified, the automatic volume control method is characterized in that it is attenuated. 5. The automatic volume control method according to any one of claims 1 to 4, wherein the weighting factor is a cosine window weighting factor. 6. A buffer circuit for extracting and accumulating an input signal for each short time interval, and an audio signal identifying circuit for identifying the type of the input signal into a target signal and a signal other than the target signal. The audio signal identification circuit includes a volume amplification circuit that performs amplification control according to the identification result, and a waveform superposition circuit that multiplies a signal of a short time section for which the volume amplification control is performed by a weighting window of the section length. And automatic volume control device.