JP2002261553A - Automatic audio gain control device, automatic audio gain control method, storage medium for storing computer program having automatic audio gain control algorithm, and computer program having automatic audio gain control algorithm - Google Patents

Abstract

(57) [Summary] [Problem] To accurately perform automatic level adjustment of an input voice even when a high-level background noise or unvoiced sound is input. SOLUTION: A voice determination means 102 for determining a property of an input voice signal is provided with a voiced sound determination function for determining the likelihood of a voiced sound of the input voice signal, and as a result of the determination, it is determined that the input voice signal is an unvoiced sound. The same process as in the case of silence is performed in the portion where the sound is made. The voiced sound determination function calculates an autocorrelation function for a predetermined type of pitch from the frame and a ratio between the autocorrelation function and the intensity of the frame, and calculates a pitch and a ratio at which the ratio of the autocorrelation function and the intensity of the frame becomes maximum. Asked,
If the maximum value exceeds a predetermined threshold, it is determined that the sound is voiced. Thus, when background noise or unvoiced sound with a large level is input, appropriate gain control can be performed without increasing the gain more than necessary.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for dividing an input audio signal into frames each having a predetermined number of samples, automatically obtaining a gain for multiplying the frames, and adjusting the level of the frames to a desired level. The present invention relates to an automatic audio gain control device, an automatic audio gain control method, a storage medium storing a computer program having an automatic audio gain control algorithm, and a computer program having an automatic audio gain control algorithm. The technology to which the present invention can be applied includes a voice recording device, voice communication, voice memo, answering machine, voice message recording device, voice recognition device, and the like.

[0002]

2. Description of the Related Art Technologies for digitizing and storing voice are used in various places such as an answering machine such as various telephones and facsimile machines, a voice message transmitting device, and a portable voice memo. In these voice recording devices, various voice encodings (voice compression techniques) may be used in order to efficiently use the memory while maintaining high quality voice.

[0003] In these voice recording devices, the level of the input voice may be a problem. When performing a message recording or message recording, a speaker performs a recording process with various voice volumes and distances from a microphone. The level at which the input voice is digitized varies depending on the characteristics of the AD converter that converts the voice to digital. In the latter case, the level can be made uniform by performing a level process in accordance with the characteristics in advance, but in the former case, it is not possible to cope with the former setting.

In order to solve the above problems, an automatic gain control (Auto G) for automatically adjusting an input voice to a desired level.
ain Control (hereinafter abbreviated as AGC) device. Some ADC devices perform AGC processing in analog before AD conversion, but in some cases, it is more efficient to perform AGC processing digitally in order to absorb the characteristics of AD conversion.

An AGC apparatus using digital processing generally determines a gain based on a ratio between an input voice intensity and a target intensity, and multiplies the gain by the gain. For example, when a DSP (Digital Signal Processor) is used as hardware, a large amount of processing occurs in division processing and square root processing, and the intensity of a frame including one or a plurality of audio samples is compared with the target intensity. A recursive method may be used in which the gain is increased when the strength of the frame is smaller than the target strength, and decreased when the strength of the frame is larger than the target strength, and the gain is used as the gain of the next frame.

In a recursive digital AGC apparatus, if the input speech signal is silent (or a relatively small background noise which is almost meaningless to a speaker or a listener), if the gain is calculated based on the intensity, an excessive Since a large gain is generated, sound detection is performed. As a technique for detecting sound, a method is used in which a predetermined threshold is given to the intensity of the frame (sum of squares of each sample in the frame), and if the intensity is equal to or smaller than the threshold, it is determined that there is no sound (or only noise). .

[0007]

However, actual voices include unvoiced sounds. Even when the intensity of a voiced sound part such as a vowel is high, since the head of a word beginning with an unvoiced sound is low in intensity, if gain control is performed based on this part, excessive gain may still occur.

Further, when gain control is performed on voices in an environment with a large background noise, such as when recording an answering machine by an outside telephone, the background noise exceeds the threshold value, and the background noise is amplified. When the threshold value is determined on the assumption of the above, the threshold value becomes high, and there is a possibility that the sound whose gain should be controlled is determined to be silent.

SUMMARY OF THE INVENTION It is an object of the present invention to accurately perform automatic level adjustment of an input voice even when a high-level background noise or unvoiced sound is input.

[0010]

According to the first aspect of the present invention,
An automatic audio gain control device for dividing an input audio signal into frames having a predetermined number of samples, automatically obtaining a gain for multiplying the frames, and adjusting a level of the frames to a desired level. And a voice determining means for determining a property of the frame, wherein the voice determining means includes a voiced sound determining means for determining whether the frame is a voiced sound or an unvoiced sound.

According to a second aspect of the present invention, in the audio automatic gain control apparatus according to the first aspect, there is provided a gain estimating means for estimating a gain by which the next frame is multiplied.
Perform different gain estimation processing when the frame is determined to be voiced and when determined to be unvoiced in the voicing sound determination unit, and further, when the frame is unvoiced in the voiced sound determination unit. When it is determined that there is no sound, the same processing is performed.

The invention described in claim 3 is the first or second invention.
In the automatic voice gain control device according to the aspect, the voiced sound determination unit uses an autocorrelation function calculation unit that calculates an autocorrelation function of a plurality of the frames for a plurality of predetermined pitches, and the plurality of autocorrelation functions. A voiced sound determining means for finally determining whether the frame is a voiced sound or an unvoiced sound is provided.

According to a fourth aspect of the present invention, in the automatic audio gain control apparatus according to the third aspect, the voiced sound determining means includes:
Determine the maximum value of the plurality of autocorrelation functions, if the maximum value of the autocorrelation function is larger than a predetermined threshold obtained for each frame, it is determined that the frame is a voiced sound, than a predetermined threshold If it is small, it is determined that the sound is unvoiced.

According to a fifth aspect of the present invention, in the audio automatic gain control apparatus according to the fourth aspect, the threshold value is determined by adding a predetermined ratio to a sum of squares of all sample points used for calculating the plurality of autocorrelation functions. It is multiplied.

According to a sixth aspect of the present invention, in the audio automatic gain control apparatus according to the third aspect, the voiced sound determining means includes:
Each of the plurality of autocorrelation functions is compared with a predetermined threshold determined for each frame, and it is determined that the frame is a voiced sound if any one of the autocorrelation functions is greater than the threshold. If it does not exist, it is determined that the sound is unvoiced.

According to a seventh aspect of the present invention, in the audio automatic gain control apparatus according to the sixth aspect, the threshold value is determined by adding a predetermined ratio to a sum of squares of all sample points used for calculating the plurality of autocorrelation functions. It is multiplied.

According to the present invention, the input speech signal is divided into frames each having a predetermined number of samples, a gain for multiplying the frames is automatically obtained, and the level of the frames is adjusted to a desired level. An automatic audio gain control method comprising an audio determining step of determining a property of the frame, wherein the audio determining step determines whether the frame is a voiced sound or an unvoiced sound. Is provided.

According to a ninth aspect of the present invention, in the voice automatic gain control method according to the eighth aspect, there is provided a gain estimating step of estimating a gain by which the next frame is multiplied.
Perform different gain estimation processing between the case where the frame is determined to be voiced and the case where it is determined to be silent in the voicing sound determination step, and further, the frame is unvoiced in the voiced sound determination step. When it is determined that there is no sound, the same processing is performed.

According to a tenth aspect of the present invention, in the automatic voice gain control method according to the eighth or ninth aspect, the voiced sound determination step calculates an autocorrelation function of a plurality of frames for a plurality of predetermined pitches. Correlation function calculation step,
And a voiced sound determination step for finally determining whether the frame is voiced or unvoiced using the plurality of autocorrelation functions.

According to an eleventh aspect of the present invention, in the voice automatic gain control method according to the tenth aspect, the voiced sound determining step determines a maximum value of the plurality of autocorrelation functions, and determines that the maximum value of the autocorrelation function is If the frame is larger than a predetermined threshold obtained for each frame, the frame is determined to be voiced, and if smaller than the predetermined threshold, the frame is determined to be unvoiced.

According to a twelfth aspect of the present invention, in the audio automatic gain control method according to the eleventh aspect, the threshold value is obtained by adding a predetermined ratio to a sum of squares of all sample points used for calculating the plurality of autocorrelation functions. It is multiplied.

According to a thirteenth aspect of the present invention, in the audio automatic gain control method according to the tenth aspect, the voiced sound determining step includes the step of determining each of the plurality of autocorrelation functions and a predetermined threshold value obtained for each frame. In comparison, if at least one of the autocorrelation functions greater than the threshold value exists, the frame is determined to be voiced, and if not, it is determined to be unvoiced.

According to a fourteenth aspect of the present invention, in the audio automatic gain control method according to the thirteenth aspect, the threshold value is obtained by adding a predetermined ratio to a sum of squares of all sample points used for calculating the plurality of autocorrelation functions. It is multiplied.

The invention according to claim 15 is installed in a computer, which divides an input audio signal into frames each having a predetermined number of samples, automatically obtains a gain by which the frames are multiplied, and obtains a level of the frames. A function to execute automatic voice gain control for the purpose of adjusting to a desired level, and a voiced determination function including a voiced sound determination function to determine whether the frame is voiced or unvoiced, A storage medium for storing a computer program having a function of executing a voice determination function for determining a property of the frame and a machine-readable algorithm for automatic gain control of voice for executing the function.

According to a sixteenth aspect of the present invention, in the computer program stored in the storage medium of the fifteenth aspect, the computer causes a computer to execute a gain estimating function for estimating a gain for multiplying a next frame. A different gain estimation process is performed between the case where the frame is determined to be voiced and the case where the frame is determined to be silent by execution of the voicing sound determination function. Is determined to be unvoiced, the same processing is performed for the case of no sound.

According to a seventeenth aspect of the present invention, in the computer program stored in the storage medium of the fifteenth or sixteenth aspect, the voiced sound determination function determines an autocorrelation function of a plurality of frames for a plurality of predetermined pitches. An autocorrelation function calculation function to calculate and a voiced sound determination function to finally determine whether the frame is voiced or unvoiced using the plurality of autocorrelation functions.

The invention according to claim 18 is the computer program stored in the storage medium according to claim 17, wherein the voiced sound determination function obtains a maximum value of the plurality of autocorrelation functions, When the maximum value is larger than a predetermined threshold value obtained for each frame, the frame is determined to be voiced, and when it is smaller than the predetermined threshold, it is determined to be unvoiced.

According to a nineteenth aspect of the present invention, in the computer program stored in the storage medium of the eighteenth aspect, the threshold value is set to a sum of squares of all sample points used for calculating the plurality of autocorrelation functions. Multiplied by the ratio of

According to a twentieth aspect of the present invention, in the computer program stored in the storage medium according to the seventeenth aspect, the voiced sound determining function is configured to determine each of the plurality of autocorrelation functions and a predetermined value determined for each frame. The frame is compared with a threshold, and if at least one of the autocorrelation functions is larger than the threshold, the frame is determined to be voiced, and if not, it is determined to be unvoiced.

According to a twenty-first aspect of the present invention, in the computer program stored in the storage medium of the twentieth aspect, the threshold value is a predetermined sum of squares of all sample points used for calculating the plurality of autocorrelation functions. Multiplied by the ratio of

The invention according to claim 22 is installed in a computer, which divides an input audio signal into frames each having a predetermined number of samples, automatically calculates a gain by which the frames are multiplied, and obtains a level of the frames. A function to execute automatic voice gain control for the purpose of adjusting to a desired level, and a voiced determination function including a voiced sound determination function to determine whether the frame is voiced or unvoiced, A computer program having a function of executing an audio determination function of determining the nature of the frame, and a machine-readable audio automatic gain control algorithm for executing the function.

According to a twenty-third aspect of the present invention, in the computer program according to the twenty-second aspect, the computer stores:
A gain estimating function for estimating a gain for multiplying the next frame is executed, and the gain estimating function is performed when the frame is determined to be voiced and when it is determined to be silent by execution of a sexual sound determination function. And a different gain estimation process is performed, and when the frame is determined to be unvoiced by the execution of the voiced sound determination function, a similar process is performed for the case of no sound.

According to a twenty-fourth aspect of the present invention, in the computer program according to the twenty-second or twenty-third aspect, the voiced sound determination function calculates an autocorrelation function of a plurality of frames for a plurality of predetermined pitches. And a voiced sound determination function that ultimately determines whether the frame is voiced or unvoiced using the plurality of autocorrelation functions.

According to a twenty-fifth aspect of the present invention, in the computer program according to the twenty-fourth aspect, the voiced sound determination function obtains a maximum value of the plurality of autocorrelation functions, and the maximum value of the autocorrelation function is determined for each frame. If the frame is larger than the predetermined threshold value, the frame is determined to be voiced, and if smaller than the predetermined threshold, the frame is determined to be unvoiced.

According to a twenty-sixth aspect of the present invention, in the computer program according to the twenty-fifth aspect, the threshold value is obtained by multiplying a sum of squares of all sample points used for calculating the plurality of autocorrelation functions by a predetermined ratio. It is.

According to a twenty-seventh aspect of the present invention, in the computer program according to the twenty-fourth aspect, the voiced sound determination function compares each of the plurality of autocorrelation functions with a predetermined threshold value obtained for each frame, If at least one of the autocorrelation functions is larger than the threshold value, the frame is determined to be voiced, and if not, it is determined to be unvoiced.

According to a twenty-eighth aspect of the present invention, in the computer program according to the twenty-seventh aspect, the threshold value is obtained by multiplying a sum of squares of all sample points used for calculating the plurality of autocorrelation functions by a predetermined ratio. It is.

[0038]

Embodiments of the present invention will be described.

In the present embodiment, a function executed by a processor of a computer according to a computer program installed in a storage medium of the computer, for example, an HDD (hard disk drive) while using a memory area such as a RAM is considered as a function. The automatic audio gain control device and the automatic audio gain control method realized by the computer will be described using a functional block diagram obtained by blocking the above and using the functional description expressed by the functional block diagram.

Further, the functional description based on such a functional block diagram also serves as a description of a computer program having an algorithm for automatic audio gain control and a storage medium storing the computer program. Here, the storage medium is not only the HDD of the computer but also a storage area such as a RAM thereof, an optical medium such as a CD-ROM, a CD-R, and a CD-RW that exists apart from the computer, or a magnetic information recording. It includes various storage media such as a medium, a communication network system for distributing a computer program to a computer.

[First Embodiment] A first embodiment of the present invention will be described with reference to FIG. FIG. 1 is a functional block diagram showing one embodiment of an audio automatic gain control (audio AGC device) to which the present invention is applied.

The voice AGC apparatus has voice input means 101
, A voice determination unit 102, a gain multiplication unit 103, and a gain estimation unit 104.

Further, the voice judging means 102 has a voiced sound judging means 201 and a voiced sound judging means 202.

Further, the sound existence determination means 201 is provided with a first
Frame intensity calculation means 301 and sound existence determination means 302
And.

Further, the voiced sound judgment means 202 includes a second frame intensity calculation means 401, an autocorrelation function calculation means 402, an autocorrelation function maximum value judgment means 403,
Voiced sound determination means 404.

Further, the gain estimation means 104
Frame intensity calculating means 501 and intensity comparing means 502
And a gain calculating means 503.

The voice input means 101 receives a digital voice signal and forms a frame of a predetermined number of samples. When the sampling frequency is 8 kHz, the number of samples is from 40 to 2 where sound is stationary.
A value in the range of 00 is desirable.

The sound determination means 102 receives the frame, and the sound determination means 201 first determines whether there is sound or no sound. In the present AGC apparatus, as a method of determining whether there is sound or no sound, the intensity of the frame (sum of squares of each sample) is calculated, and if the value is larger than a predetermined threshold, the frame is determined to be sound. If it is small, use a method of silence. That is, in the sound determination unit 201, first, the first frame intensity calculation unit 301 calculates the intensity (1) of the frame. Further, the sound determination means 302 compares the intensity (1) of the frame with a predetermined threshold, and determines that the frame is sound if the intensity (2) of the frame is higher than the threshold. Is smaller, it is determined that there is no sound.

The voiced sound judging means 202 judges whether the frame judged as voiced by the voiced sound judging means 201 is a voiced sound or an unvoiced sound. In the present AGC apparatus, as a method of determining whether a voiced sound or an unvoiced sound, the autocorrelation function of a frame for a plurality of predetermined pitches is calculated, and the ratio between the maximum value of the autocorrelation function and the frame intensity is obtained. When the ratio is larger than a predetermined value, a method of making a voiced sound is used.

In the AGC apparatus, even in the case of a frame determined to be voiced, by processing other than the voiced frame in the same manner as the unvoiced frame, when the background noise level is high or in the case of the unvoiced frame. Is not multiplied by an unnecessarily high gain.

That is, in the voiced sound judging means 202, first, the intensity (2) of the frame is calculated by the second frame intensity calculating means 401. Since the autocorrelation function is calculated using not only each sample of the current frame but also the samples of the previous frame, the intensity (2) of this frame is calculated as the intensity of all the samples used in the autocorrelation function calculation. Since this is a calculation, the value is different from the intensity (1) of the frame.

When the number of samples for calculating the intensity (2) of the frame is an integral multiple of the number of samples for calculating the intensity (1) of the frame (that is, the number of samples of the frame), the past and present frames are used. By adding the intensities (1) of the frames, the intensity (2) of the frame can be obtained, and the number of product-sum calculations can be reduced.

Further, the autocorrelation function calculating means 402 calculates an autocorrelation function for a plurality of predetermined pitches. The autocorrelation function is a technique used for estimating the pitch of the frame, and its calculation method is well known to those skilled in the art. The pitch for calculating the autocorrelation function is 50 Hz to 4 which is the pitch range of human voice.
In the case of 00 Hz, that is, a sampling frequency of 8 kHz, it is preferable to use about 20 to 120 samples. When the maximum value of the autocorrelation function (peak of the autocorrelation function sequence) in this range is clearly obtained, the frame can be considered to be a voiced sound, and the given pitch is the pitch of the frame. Conceivable.

Further, the autocorrelation function maximum value judging means 40
In 3, the maximum value is determined from the plurality of autocorrelation functions determined for each pitch.

Further, in the voiced sound determination means 404,
A ratio between the maximum value of the autocorrelation function and the intensity (2) of the frame is obtained, and the ratio is compared with a predetermined threshold. If the ratio is larger than the threshold, it is determined that the sound is a voiced sound. If it is smaller than the threshold, it is determined that the sound is unvoiced.

Also, instead of obtaining the ratio between the maximum value of the autocorrelation function and the intensity (2) of the frame and comparing it with a threshold value of a predetermined ratio, the intensity of the frame (2) is determined in advance by a predetermined ratio. May be determined as a voiced sound when the maximum value of the autocorrelation function is large with respect to the multiplied value, and may be determined as an unvoiced sound when the maximum value is small.

Further, in the gain multiplying means 103,
The frame is multiplied by the gain determined by the gain estimator 104 in the previous frame. If the current frame is the first processing frame, a predetermined initial value may be used as the gain. The initial value of the gain is 1, for example.

The frame multiplied by the gain is output as an AGC output and input to an application such as speech coding and speech recognition.

Further, in the gain estimating means 104,
On the basis of the determination result in the voice determination means 102,
In the case of a silent frame and an unvoiced sound frame (hereinafter referred to as a silent / unvoiced frame) and in the case of a voiced sound frame, unique processing is performed for each of them, and gain estimation for the next frame is performed. As a method of gain estimation, a method known to those skilled in the art can be adopted. The method is disclosed in
This is like the method described in Japanese Patent Publication No. 214940. Therefore, an example is described below, but is not limited thereto.

First, for the silent / unvoiced frame,
The gain calculator 503 does not change the gain. However, when a predetermined number of silent / unvoiced sound frames continue, the gain is reset to the initial value.

For voiced sound frames, first,
In the third frame strength calculating means 501, the strength of the frame multiplied by the gain is calculated, and the strength (3) of the frame is obtained. The strength (3) of the frame differs from the strength (1) of the frame and the strength (2) of the frame in terms of the strength of the frame multiplied by the gain.

Further, the intensity comparing means 502 compares the intensity (3) of the frame with a predetermined intensity target value.

Further, in the gain calculating means 503, when the strength (3) of the frame is larger than a predetermined strength target value in the strength comparing means, the gain is reduced by a predetermined value. Increase by a predetermined value.

[Second Embodiment] A second embodiment of the present invention will be described with reference to FIG. FIG. 2 is a functional block diagram showing an embodiment of an automatic audio gain control (audio AGC device) to which the present invention is applied. The same parts as those in the first embodiment are denoted by the same reference numerals, and the description is omitted.

The voice AGC apparatus is provided with voice input means 101
, A voice determination unit 102, a gain multiplication unit 103, and a gain estimation unit 104.

Further, the voice judging means 102 has a voiced sound judging means 201 and a voiced sound judging means 202.

Further, the sound existence determination means 201 includes a first
Frame intensity calculation means 301 and sound existence determination means 302
And.

Further, the voiced sound determining means 202 includes a second frame intensity calculating means 401, an autocorrelation function calculating means 402, and a voiced sound determining means 403.

Further, the gain estimating means 104
Frame intensity calculating means 501 and intensity comparing means 502
And a gain calculating means 503.

The present AGC apparatus has an AGC
Only the operation of the voiced sound determination means 202 is shown because only the voiced sound determination means 202 differs from the apparatus.

In the voiced sound judging means 202, first, the intensity (2) of the frame is calculated by the second frame intensity calculating means 401.

If the number of samples for calculating the intensity (2) of the frame is an integral multiple of the number of samples for calculating the intensity (1) of the frame (that is, the number of samples of the frame), the past and current frames are calculated. By adding the intensities (1) of the frames, the intensity (2) of the frame can be obtained, and the number of product-sum calculations can be reduced.

Further, the autocorrelation function calculating means 402 calculates an autocorrelation function for a plurality of predetermined pitches. The autocorrelation function is a technique used for estimating the pitch of the frame, and its calculation method is well known to those skilled in the art. The pitch for calculating the autocorrelation function is 50 Hz to 4 which is the pitch range of human voice.
In the case of 00 Hz, that is, a sampling frequency of 8 kHz, it is preferable to use about 20 to 120 samples. When the maximum value of the autocorrelation function (peak of the autocorrelation function sequence) in this range is clearly obtained, the frame can be considered to be a voiced sound, and the given pitch is the pitch of the frame. Conceivable.

Further, in the voiced sound determination means 403,
A ratio between each of the plurality of obtained autocorrelation functions and the intensity (2) of the frame is obtained, and this ratio is compared with a predetermined threshold value. It is determined to be a voice sound, and if there is no voice sound, it is determined to be a voiceless sound.

Also, instead of obtaining the ratio between the autocorrelation function and the intensity (2) of the frame and comparing it with a threshold value of a predetermined ratio, a threshold value of a predetermined ratio is previously assigned to the intensity (2) of the frame. If there is at least one autocorrelation function larger than the multiplied value, the voiced sound may be determined, and if not, the voiced sound may be determined to be unvoiced.

[0076]

According to the first aspect of the present invention, an input audio signal is divided into frames having a predetermined number of samples, a gain for multiplying the frames is automatically obtained, and the level of the frames is adjusted to a desired level. An automatic voice gain control device, comprising: voice determining means for determining a property of the frame, wherein the voice determining means determines whether the frame is a voiced sound or an unvoiced sound. Since the determination means is provided, if the result of the determination by the voiced sound determination means is determined to be unvoiced, by performing the same processing as in the case of no sound, a large level of background noise or unvoiced sound is input. Thus, appropriate gain control can be performed without increasing the gain more than necessary.

According to a second aspect of the present invention, in the automatic voice gain control apparatus according to the first aspect, there is provided a gain estimating means for estimating a gain by which the next frame is to be multiplied.
Perform different gain estimation processing when the frame is determined to be voiced and when determined to be unvoiced in the voicing sound determination unit, and further, when the frame is unvoiced in the voiced sound determination unit. When it is determined that no sound is present, the same processing is performed in the case of no sound, so that the audio automatic gain control device according to claim 1 can be realized with a simpler configuration.

The third aspect of the present invention is the first or second aspect.
In the automatic voice gain control device according to the aspect, the voiced sound determination unit uses an autocorrelation function calculation unit that calculates an autocorrelation function of a plurality of the frames for a plurality of predetermined pitches, and the plurality of autocorrelation functions. Since the voiced sound determining means for finally determining whether the frame is a voiced sound or an unvoiced sound is provided, the automatic voice gain control device according to claims 1 and 2 can be realized with a simpler configuration.

According to a fourth aspect of the present invention, in the audio automatic gain control apparatus according to the third aspect, the voiced sound determining means is
Determine the maximum value of the plurality of autocorrelation functions, if the maximum value of the autocorrelation function is larger than a predetermined threshold obtained for each frame, it is determined that the frame is a voiced sound, than a predetermined threshold If the volume is small, it is determined that the voice is unvoiced, so that the voice automatic gain control device according to claim 3 can be realized with a simpler configuration.

According to a fifth aspect of the present invention, in the audio automatic gain control apparatus according to the fourth aspect, the threshold value is obtained by adding a predetermined ratio to a sum of squares of all sample points used for calculating the plurality of autocorrelation functions. Since the multiplication is performed, the audio automatic gain control device according to claim 4 can be realized with a simpler configuration.

According to a sixth aspect of the present invention, in the audio automatic gain control apparatus according to the third aspect, the voiced sound determining means includes:
Each of the plurality of autocorrelation functions is compared with a predetermined threshold determined for each frame, and it is determined that the frame is a voiced sound if any one of the autocorrelation functions is greater than the threshold. If it does not exist, it is determined that the sound is unvoiced, so that the voice automatic gain control device according to claim 3 can be realized with a simpler configuration.

According to a seventh aspect of the present invention, in the automatic audio gain control apparatus according to the sixth aspect, the threshold value is obtained by adding a predetermined ratio to a sum of squares of all sample points used for calculating the plurality of autocorrelation functions. Since the multiplication is performed, the audio automatic gain control device according to claim 6 can be realized with a simpler configuration.

According to the present invention, the input audio signal is divided into frames each having a predetermined number of samples, a gain for multiplying the frames is automatically obtained, and the level of the frames is adjusted to a desired level. An automatic audio gain control method comprising an audio determining step of determining a property of the frame, wherein the audio determining step determines whether the frame is a voiced sound or an unvoiced sound. Therefore, when the voiced sound determination step determines that the sound is unvoiced, the subsequent processing is performed in the same manner as in the case of no sound, so that a large level of background noise or unvoiced sound is input. As described above, appropriate gain control can be performed without increasing the gain.

According to a ninth aspect of the present invention, in the automatic voice gain control method of the eighth aspect, there is provided a gain estimating step of estimating a gain by which the next frame is to be multiplied.
Perform different gain estimation processing between the case where the frame is determined to be voiced and the case where it is determined to be silent in the voicing sound determination step, and further, the frame is unvoiced in the voiced sound determination step. If it is determined that no sound is present, the same processing is performed for the case of no sound, so that the automatic audio gain control method according to claim 8 can be realized with a simpler configuration.

According to a tenth aspect of the present invention, in the voice automatic gain control method according to the eighth or ninth aspect, the voiced sound determination step calculates an autocorrelation function of a plurality of frames for a plurality of predetermined pitches. Correlation function calculation step,
And a voiced sound determining step of finally determining whether the frame is a voiced sound or an unvoiced sound using the plurality of autocorrelation functions, so that the automatic voice processing according to claim 8 or 9 has a simpler configuration. A gain control method can be realized.

According to an eleventh aspect of the present invention, in the voice automatic gain control method according to the tenth aspect, the voiced sound determining step obtains a maximum value of the plurality of autocorrelation functions, and determines a maximum value of the autocorrelation function. The method according to claim 10, wherein the frame is determined to be a voiced sound if it is larger than a predetermined threshold value obtained for each frame, and is determined to be an unvoiced sound if it is smaller than the predetermined threshold value. Can be realized.

According to a twelfth aspect of the present invention, in the audio automatic gain control method according to the eleventh aspect, the threshold value is obtained by adding a predetermined ratio to a sum of squares of all sample points used for calculating the plurality of autocorrelation functions. Since the multiplication is performed, the audio automatic gain control method according to claim 11 can be realized with a simpler configuration.

According to a thirteenth aspect of the present invention, in the voice automatic gain control method according to the tenth aspect, the voiced sound determination step includes the step of determining each of the plurality of autocorrelation functions and a predetermined threshold value obtained for each frame. 11. A simpler configuration, since the frame is determined to be voiced if at least one of the autocorrelation functions is larger than the threshold, and the frame is determined to be unvoiced if not. The described automatic audio gain control method can be realized.

According to a fourteenth aspect of the present invention, in the audio automatic gain control method according to the thirteenth aspect, the threshold value is obtained by adding a predetermined ratio to a sum of squares of all sample points used for calculating the plurality of autocorrelation functions. Since the multiplication is performed, the audio automatic gain control method according to claim 13 can be realized with a simpler configuration.

According to a fifteenth aspect of the present invention, there is provided a storage medium for storing a computer program having an algorithm for machine-readable audio automatic gain control, wherein the computer program is installed in a computer,
This computer has an automatic audio gain control for dividing an input audio signal into frames having a predetermined number of samples, automatically obtaining a gain for multiplying the frames, and adjusting a level of the frames to a desired level. And a voiced determination function including a voiced sound determination function to determine whether the frame is a voiced sound or unvoiced sound, a function to perform a voice determination function to determine the nature of the frame, Therefore, if the result of the determination in the voiced sound determination step is determined to be unvoiced, by performing subsequent processing similar to the case of silence, when a large level of background noise or unvoiced sound is input, Appropriate gain control can be performed without increasing the gain more than necessary.

According to a sixteenth aspect of the present invention, in the computer program stored in the storage medium of the fifteenth aspect, the computer causes a computer to execute a gain estimating function for estimating a gain for multiplying a next frame. A different gain estimation process is performed between the case where the frame is determined to be voiced and the case where the frame is determined to be silent by execution of the voicing sound determination function. Is determined to be unvoiced, the same processing is performed in the case of no sound, so that the computer program according to claim 15 can be realized with a simpler configuration.

According to a seventeenth aspect of the present invention, in the computer program stored in the storage medium according to the fifteenth or sixteenth aspect, the voiced sound determination function determines an autocorrelation function of a plurality of frames for a plurality of predetermined pitches. An autocorrelation function calculation function for calculating, and a voiced sound determination function for finally determining whether the frame is a voiced sound or an unvoiced sound using the plurality of autocorrelation functions, so that a simpler configuration is provided. Thus, the computer program according to claim 15 or 16 can be realized.

According to an eighteenth aspect of the present invention, in the computer program stored in the storage medium of the seventeenth aspect, the voiced sound determining function obtains a maximum value of the plurality of autocorrelation functions, If the maximum value is larger than a predetermined threshold value obtained for each frame, the frame is determined to be voiced, and if smaller than the predetermined threshold, it is determined to be unvoiced sound. Item 17 can be realized.

According to a nineteenth aspect of the present invention, in the computer program stored in the storage medium of the eighteenth aspect, the threshold value is set to a sum of squares of all sample points used for calculating the plurality of autocorrelation functions. Therefore, the computer program according to claim 18 can be realized with a simpler configuration.

According to a twentieth aspect of the present invention, in the computer program stored in the storage medium according to the seventeenth aspect, the voiced sound determination function includes a step of determining each of the plurality of autocorrelation functions and a predetermined function determined for each frame. Compare with a threshold, if at least one of the autocorrelation function is larger than the threshold is determined, the frame is determined to be a voiced sound, if not present is determined to be unvoiced sound, with a simpler configuration A computer program according to claim 17 can be realized.

According to a twenty-first aspect of the present invention, in the computer program stored in the storage medium of the twentieth aspect, the threshold value is a predetermined sum of squares of all sample points used for calculating the plurality of autocorrelation functions. Therefore, the computer program according to claim 20 can be realized with a simpler configuration.

According to a twenty-second aspect of the present invention, there is provided a computer program having an algorithm for machine-readable audio automatic gain control.
Installed on a computer, the computer divides the input audio signal into frames of a predetermined number of samples, and automatically determines a gain for multiplying the frames;
A voiced determination function including a function of performing automatic voice gain control for the purpose of adjusting the level of the frame to a desired level and a voiced sound determination function of determining whether the frame is voiced or unvoiced And a function of executing a voice determination function of determining the nature of the frame, so that when the voiced voice is determined to be unvoiced as a result of the determination in the voiced voice determination step, the subsequent By performing the process, when a high-level background noise or unvoiced sound is input, appropriate gain control can be performed without increasing the gain more than necessary.

The invention according to claim 23 is the computer program according to claim 22, wherein
A gain estimating function for estimating a gain for multiplying the next frame is executed, and the gain estimating function is performed when the frame is determined to be voiced and when it is determined to be silent by execution of a sexual sound determination function. 23. A simpler configuration according to claim 22, wherein a different gain estimation process is performed in accordance with the above, and further, when the frame is determined to be unvoiced by the execution of the voiced sound determination function, a similar process is performed in the case of silence. A computer program can be realized.

According to a twenty-fourth aspect of the present invention, in the computer program according to the twenty-second or twenty-third aspect, the voiced sound determination function calculates an autocorrelation function for calculating a plurality of autocorrelation functions of a plurality of frames for a plurality of predetermined pitches. 24. A simpler configuration, comprising a function and a voiced sound determination function for finally determining whether the frame is voiced or unvoiced using the plurality of autocorrelation functions. The described computer program can be realized.

According to a twenty-fifth aspect of the present invention, in the computer program according to the twenty-fourth aspect, the voiced sound determination function obtains a maximum value of the plurality of autocorrelation functions, and the maximum value of the autocorrelation function is determined for each frame. 25. The computer program according to claim 24, wherein the frame is determined to be a voiced sound if the frame is larger than a predetermined threshold value, and an unvoiced sound is determined if the frame is smaller than the predetermined threshold value. Therefore, the computer program according to claim 24 can be realized with a simpler configuration.

The invention according to claim 26 is the computer program according to claim 25, wherein the threshold value is obtained by multiplying a sum of squares of all sample points used for calculating the plurality of autocorrelation functions by a predetermined ratio. Therefore, the computer program according to claim 25 can be realized with a simpler configuration.

According to a twenty-seventh aspect of the present invention, in the computer program according to the twenty-fourth aspect, the voiced sound determining function compares each of the plurality of autocorrelation functions with a predetermined threshold value obtained for each frame, 25. The computer according to claim 24, wherein the frame is determined to be a voiced sound if at least one of the autocorrelation functions greater than the threshold value is present, and is determined to be an unvoiced sound if not. A program can be realized.

According to a twenty-eighth aspect of the present invention, in the computer program according to the twenty-seventh aspect, the threshold value is obtained by multiplying a sum of squares of all sample points used for calculating the plurality of autocorrelation functions by a predetermined ratio. Therefore, the computer program according to claim 27 can be realized with a simpler configuration.

[Brief description of the drawings]

FIG. 1 is a functional block diagram of an automatic audio gain device (AGC device) according to a first embodiment of the present invention.

FIG. 2 is a functional block diagram of an automatic audio gain device (AGC device) according to a second embodiment of the present invention.

[Explanation of symbols]

 Reference Signs List 102 voice determination means (step, function) 202 voiced sound determination means (step, function) 104 gain estimation means (step, function) 402 autocorrelation function calculation means (step, function) 404 voiced sound determination means (step, function)

 ──────────────────────────────────────────────────続 き Continuing from the front page (54) [Title of the Invention] A storage medium for storing a computer program having an automatic audio gain control device, an automatic audio gain control method, an algorithm for automatic audio gain control, and an automatic audio gain control Computer program with different algorithms

Claims (28)

[Claims] 1. An automatic audio gain for dividing an input audio signal into frames each having a predetermined number of samples, automatically obtaining a gain for multiplying the frames, and adjusting a level of the frames to a desired level. The control device, further comprising: a voice determination unit that determines a property of the frame, wherein the voice determination unit includes a voiced sound determination unit that determines whether the frame is a voiced sound or an unvoiced sound. Automatic gain control device. 2. A gain estimating means for estimating a gain by which a next frame is multiplied, wherein the gain estimating means determines whether the frame is a sound by the sexual sound determining means and determines that the frame is silent. 2. A gain estimation process different from that performed when the frame is determined to be unvoiced by the voiced sound determination means, and a similar process is performed when the frame is unvoiced. Audio automatic gain control device. 3. An autocorrelation function calculation means for calculating an autocorrelation function of a plurality of frames for a plurality of predetermined pitches, and said voiced sound determination means includes a voiced sound using the plurality of autocorrelation functions. 3. The automatic voice gain control device according to claim 1, further comprising voiced sound determination means for finally determining whether the voice is a voice or a voiceless voice. 4. The voiced sound determination means obtains a maximum value of the plurality of autocorrelation functions, and if the maximum value of the autocorrelation function is larger than a predetermined threshold value obtained for each frame, the voiced sound determination means 4. The automatic voice gain control device according to claim 3, wherein the voice is determined to be a voice, and if the voice is smaller than a predetermined threshold, it is determined to be an unvoiced voice. 5. The automatic audio gain according to claim 4, wherein the threshold value is obtained by multiplying a sum of squares of all sample points used for calculating the plurality of autocorrelation functions by a predetermined ratio. Control device. 6. The voiced sound determining means compares each of the plurality of autocorrelation functions with a predetermined threshold value obtained for each frame, and if any one of the autocorrelation functions is larger than the threshold value. 4. The automatic audio gain control device according to claim 3, wherein if the frame is determined to be a voiced sound, if not, the frame is determined to be an unvoiced sound. 7. The automatic audio gain according to claim 6, wherein the threshold is obtained by multiplying a sum of squares of all sample points used for calculating the plurality of autocorrelation functions by a predetermined ratio. Control device. 8. An automatic audio gain for dividing an input audio signal into frames having a predetermined number of samples, automatically obtaining a gain for multiplying the frames, and adjusting a level of the frames to a desired level. A control method, comprising: a voice determination step of determining a property of the frame; wherein the voice determination step includes a voiced sound determination step of determining whether the frame is a voiced sound or an unvoiced sound. Automatic gain control method. 9. A gain estimating step of estimating a gain by which a next frame is multiplied, wherein the gain estimating step is performed when the frame is determined to be voiced in the sexual sound determination step and when the frame is determined to be silent. 9. A gain estimation process different from that performed when the frame is determined, and the same process is performed when the frame is determined to be unvoiced in the voiced sound determination step when the frame is unvoiced. Voice automatic gain control method. 10. The voiced sound determination step includes: an autocorrelation function calculation step of calculating an autocorrelation function of a plurality of the frames for a plurality of predetermined pitches; and the voiced sound is determined using the plurality of autocorrelation functions. 10. The voice automatic gain control method according to claim 8, further comprising a voiced sound determination step of finally determining whether the sound is unvoiced or not. 11. The voiced sound determination step includes determining a maximum value of the plurality of autocorrelation functions, and determining that the frame has a value if the maximum value of the autocorrelation function is larger than a predetermined threshold value determined for each frame. 11. The automatic voice gain control method according to claim 10, wherein it is determined that the sound is a vocal sound, and when it is smaller than a predetermined threshold value, it is determined that the sound is an unvoiced sound. 12. The automatic audio gain according to claim 11, wherein the threshold value is obtained by multiplying a sum of squares of all sample points used for calculating the plurality of autocorrelation functions by a predetermined ratio. Control method. 13. The voiced sound determination step includes comparing each of the plurality of autocorrelation functions with a predetermined threshold value obtained for each frame, and if any one of the autocorrelation functions is larger than the threshold value. 11. The automatic audio gain control method according to claim 10, wherein if the frame is determined to be a voiced sound, if not, the frame is determined to be an unvoiced sound. 14. The automatic audio gain according to claim 13, wherein the threshold value is obtained by multiplying a sum of squares of all sample points used for calculating the plurality of autocorrelation functions by a predetermined ratio. Control method. 15. Installed in a computer, the computer divides an input audio signal into frames each having a predetermined number of samples, automatically determines a gain for multiplying the frames, and adjusts the level of the frames to a desired level. And a voiced determination function including a voiced sound determination function for determining whether the frame is a voiced sound or an unvoiced sound, wherein the function of the frame is determined. A storage medium for storing a computer program having a function of executing a voice determination function to execute the function, and a machine-readable algorithm for automatic gain control of a machine to execute the function. 16. A computer for executing a gain estimating function for estimating a gain to be multiplied by a next frame, wherein the gain estimating function is performed when the frame is determined to be voiced by execution of a sexual sound determination function. Perform a different gain estimation process in the case where it is determined to be unvoiced, and further perform the same process in the case of silence when the frame is determined to be unvoiced by the execution of the voiced sound determination function. A storage medium for storing the computer program according to claim 15, wherein: 17. The voiced sound determination function includes: an autocorrelation function calculation function of calculating an autocorrelation function of a plurality of frames for a plurality of predetermined pitches; and a voiced sound generated by using the plurality of autocorrelation functions. 17. A voiced sound determination function for finally determining whether the sound is a voiced sound or an unvoiced sound.
A storage medium for storing the computer program described above. 18. The voiced sound determination function obtains a maximum value of the plurality of autocorrelation functions, and if the maximum value of the autocorrelation function is larger than a predetermined threshold obtained for each frame, the voiced sound determination function determines The storage medium for storing a computer program according to claim 17, wherein the computer program is determined to be a vocal sound, and is determined to be an unvoiced sound if it is smaller than a predetermined threshold. 19. The computer program according to claim 18, wherein the threshold value is obtained by multiplying a sum of squares of all sample points used for calculating the plurality of autocorrelation functions by a predetermined ratio. Storage media to store. 20. The voiced sound determination function compares each of the plurality of autocorrelation functions with a predetermined threshold value obtained for each frame, and determines whether at least one of the autocorrelation functions is larger than the threshold value. 18. The storage medium storing the computer program according to claim 17, wherein if the frame is determined to be a voiced sound, and if not, the frame is determined to be an unvoiced sound. 21. The computer program according to claim 20, wherein the threshold value is obtained by multiplying a sum of squares of all sample points used for calculating the plurality of autocorrelation functions by a predetermined ratio. Storage media to store. 22. Installed in a computer, the input audio signal is divided into frames each having a predetermined number of samples, a gain for multiplying the frames is automatically obtained, and the level of the frames is adjusted to a desired level. And a voiced judgment function including a voiced sound judgment function for judging whether the frame is a voiced sound or an unvoiced sound, wherein the property of the frame is judged. A computer program having a function of executing a voice determination function to execute, and a machine-readable voice automatic gain control algorithm for executing the function. 23. A computer that executes a gain estimating function for estimating a gain by which a next frame is to be multiplied, wherein the gain estimating function includes a case where the frame is determined to be sound by execution of a sexual sound determination function. Perform a different gain estimation process in the case where it is determined to be unvoiced, and further perform the same process in the case of silence when the frame is determined to be unvoiced by the execution of the voiced sound determination function. The computer program according to claim 22, characterized in that: 24. The voiced sound determination function includes: an autocorrelation function calculation function for calculating an autocorrelation function of a plurality of the frames for a plurality of predetermined pitches; and a voiced sound using the plurality of autocorrelation functions. 24. A voiced sound determination function for finally determining whether the sound is a voiced sound or an unvoiced sound.
Computer program as described. 25. The voiced sound determination function obtains a maximum value of the plurality of autocorrelation functions, and if the maximum value of the autocorrelation function is larger than a predetermined threshold obtained for each frame, the voiced sound determination function 25. The computer program according to claim 24, wherein the computer program is determined to be a vocal sound, and is determined to be unvoiced if it is smaller than a predetermined threshold. 26. The computer program according to claim 25, wherein the threshold value is obtained by multiplying a sum of squares of all sample points used for calculating the plurality of autocorrelation functions by a predetermined ratio. 27. The voiced sound determination function compares each of the plurality of autocorrelation functions with a predetermined threshold value obtained for each frame, and if any one of the autocorrelation functions is larger than the threshold value. 25. The computer program according to claim 24, wherein if the frame is determined to be voiced, the frame is determined to be unvoiced if not present. 28. The computer program according to claim 27, wherein the threshold value is obtained by multiplying a sum of squares of all sample points used for calculating the plurality of autocorrelation functions by a predetermined ratio.