JP2910417B2 - Voice music discrimination device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention automatically determines whether a continuously input signal used as a pre-processing device such as an audio device is a voice or another signal (music or the like). The present invention relates to an audio / music discriminating apparatus.

[0002]

2. Description of the Related Art Recent stereo apparatuses, televisions, and the like are equipped with functions such as surround for creating sound effects. While these are very effective for sources such as music, they are less clear when used mainly for audio such as news programs. Therefore, if it is possible to automatically determine whether the source of the television or radio is mainly of voice or other, it is possible to optimally control the sound field and the frequency characteristics according to the result, and it becomes very easy to hear.

[0003] The conventional method of sound and music discrimination utilizes that an input signal is a stereo signal. In the case of a music source, the left (L) channel and right (R) channel signals are independent of each other, and the correlation between the two channels is low. Conversely, in a source mainly composed of audio such as a news program, the L signal and the R signal are almost the same, and the correlation between the two channels is high. Then, the difference between the amplitudes of the L signal and the R signal is calculated,
When the difference is large, it can be determined as music, and when the difference is small, it can be determined as voice. Further, the correlation value between the L signal and the R signal can be calculated, and when the correlation value is large, it can be determined that the voice is sound, and when the correlation value is small, it can be determined that music.

[0004]

However, the above-described audio / music discriminating apparatus is effective for a stereo source, but cannot be discriminated for a monaural source having no difference between an L signal and an R signal. Have.

An object of the present invention is to provide a voice / music discriminating apparatus which can accurately discriminate between voice and music without discriminating between a monaural signal and a stereo signal. It is assumed that.

[0006]

In order to solve the above-mentioned problems, a voice / music discriminating apparatus according to the present invention comprises a power calculating unit,
A sound determination unit, a voice music determination device including a determination unit,
The power calculator divides the input signal into frames of a fixed time.
And calculate the sound power for each frame,
The determining unit determines that the acoustic power calculated by the power calculating unit is an acoustic power.
Sound threshold if it is greater than the power threshold, and no sound if it is lower
The section is determined, and the determination unit determines the first and second intervals for each of a plurality of frames.
2 and apply to the first and second determinations.
If not, the previous determination is made the current determination and the first
The judgment is that the number of frames judged as a silent section is greater than the audio threshold
Is larger, the plurality of frames are determined to be audio, and the second
The judgment is that the number of frames judged as a sound section is greater than the music threshold
Is large, it is determined that multiple frames are music.
Features. Further, the audio / music discriminating apparatus of the present invention
-Sound / Music judgment with calculation unit, sound / silence judgment unit, judgment unit
In another apparatus, the power calculation unit outputs the input signal for a predetermined time.
And the sound power of each frame
The sound / no-sound determining unit calculates the sound calculated by the power calculating unit.
Sound zone if the sound power is greater than the first sound power threshold
If the sound power is smaller than the second sound power threshold,
And the determining unit performs the first and second determinations for each of a plurality of frames.
And do not apply to the first and second judgments
In this case, the previous determination is the current determination, and the first determination is
The number of frames determined as silent sections is greater than the audio threshold
In this case, a plurality of frames are determined to be audio, and the second determination is
The number of frames judged as a sound section is larger than the music threshold
In this case, a plurality of frames are determined to be music.
You.

Further , the voice / music discriminating apparatus of the present invention
-Sound / Music judgment with calculation unit, sound / silence judgment unit, judgment unit
In another apparatus, the power calculation unit outputs the input signal for a predetermined time.
And the sound power of each frame
The sound / no-sound determining unit calculates the sound calculated by the power calculating unit.
Sound interval when the sound power is greater than the sound power threshold,
If it is smaller, it is determined to be a silent section, and the determination unit
The first and second determinations are made for each of a plurality of frames and a second plurality of frames.
And does not apply to the first and second judgments
In this case, the previous determination is made the current determination and the first determination
Indicates a frame determined to be a silent section in the first plurality of frames.
The number of frames is greater than the audio threshold and
And the number of frames judged as silent sections is larger than the audio threshold
If the second frame is determined to be speech, the second
Judgment was determined to be a sound section in the first plurality of frames.
If the number of frames is greater than the music threshold and the second
The number of frames determined to be sounded in
Is larger, the second plural frames are determined to be music.
And features. In addition, the audio / music discriminating apparatus of the present invention includes:
Power calculation unit, sound / non-speech judgment unit, voice sound with judgment unit
In a music discriminating apparatus, a power calculation unit controls an input signal to be constant.
In addition to dividing into time frames, the sound power of each frame
Is calculated, and the sound / silence determination section calculates
If the sound power is greater than the first sound power threshold.
Sound section, silent section when smaller than the second sound power threshold
And the determination unit determines that the first plurality of consecutive frames and the second
The first and second determinations are made for each of a plurality of frames, and the first
And if not applicable to the second determination, the previous determination
Is the current determination, and the first determination is for the first plurality of frames.
The number of frames determined to be a silent section in
Large, determined to be a silent section in the second multiple frames
If the number of frames is greater than the audio threshold, the second
The frame is determined to be voice, and the second determination is performed by the first multiple frames.
The number of frames determined to be sound in the
Is judged to be a sound section in the second plural frames.
If the number of frames obtained is larger than the music threshold, the second
It is characterized in that several frames are determined to be music.

[0008]

According to the present invention, the above-mentioned configuration makes use of the fact that a silence section always exists in continuously uttered speech but hardly exists in music, and a plurality of fixed frames are used for the input signal. -Voice and music can be distinguished with high accuracy by determining voice and music based on the existence ratio of sound and silence between music and music. Is output, the determination result is fixed to the dominant side in the overall flow, and a voice / music determination device with less erroneous determination can be provided.

Further, according to the present invention, a more accurate judgment can be realized by outputting the judgment result of the voice music only when the same judgment result is continuously obtained, so that a smooth judgment of the voice music can be realized. A switchable voice / music discriminating apparatus can be provided.

[0010]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An audio / music discriminating apparatus according to an embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram of a voice / music discriminating apparatus according to an embodiment of the present invention. (FIG. 1)
Is a power calculation unit for calculating the power of the input signal, and 2 is a sound / silence determination unit for comparing the input signal of the frame with sound or silence by comparing with a threshold value of the power. is there. Numeral 3 denotes a judgment unit for judging voice music between a plurality of frames, and outputting the judgment result in the current frame based on the previous judgment result. FIG. 2 is a main part flowchart for explaining the operation of the judging section 3.

Next, the operation of the voice discriminating apparatus in the above embodiment will be described in detail with reference to FIG. Here, the input signal is a stereo signal considering audio equipment, a television, and the like as targets. The input L signal and R signal of the stereo signal are mixed and input to the power calculation unit 1 as L + R. The power calculation unit 1 calculates the cumulative value or the average value of the amplitude of the section at intervals of a predetermined time (frame) as a power value in the frame. The sound / non-speech judging section 2 judges the sound / non-speech of each frame using the obtained power value. Here the current frame - determines if the value P, and the threshold value of the voice activity detection and P _t, determines that sound when satisfying equation (1), if not satisfied with the silence - power in beam.

[0013]

(Equation 1)

[0014] In addition, the sound / silence judgment is made by setting the judgment threshold to P
_t1 , _Pt2 (provided that _Pt2 is larger than _Pt1 )
Is determined as silence when (Equation 2) is satisfied,
When (Equation 3) is satisfied, the sound may be determined.

[0015]

(Equation 2)

[0016]

(Equation 3)

Using the result of the sound / non-speech judgment for each frame, the judgment unit 3 judges the sound music for each unit larger than the frame. Hereinafter, the operation of the determination unit 3 will be described in detail along the flowchart of FIG.

Hereinafter, it is assumed that the voice / music judgment is performed for each F frame. The determination interval F may be set to a value that includes an average of three or four syllables in the case of continuously uttered speech. Actually, if the value is set to a value between 1 second and 2 seconds, a silent portion will be included in the voice at a certain ratio, and the accuracy of voice / music determination can be improved.

First, based on the result of the sound / non-speech discriminating section 2, in steps 21 and 22, if there is no sound, the silence counter is incremented, and if there is sound, the sound counter is incremented. Next, the counter of the number of frames is incremented (step 23). If the sound / silence determination for the F frame has been completed in the determination of step 24, the frame counter is reset (step 25). . Here, let L _{T be} the threshold value for voice determination, and let H _T be the threshold value for music determination.

In step 26, the silence counter value and L _T
If the silence counter is large, it is determined that the input signal is a voice at the present time, and in step 27, the determination flag is turned on and the information is output to the outside. Even when the silence state is continuous, the silence is considered to be a kind of voice and is determined on the voice side. Here, when the judgment flag is “1”, the sound is
When it is “0”, it is music. If towards the silence counter value is smaller at step 26, compares the voice counter value and H _T in step 28, if it is greater, it determines that the input signal is music, the turn off determination flag in step 29 Output information to the outside. If it is smaller, it is difficult to determine, so the state of the flag that has already been set is kept as it is. Thereafter, the processing proceeds to the processing of the next frame, and the same processing is repeated.

The discriminating section 3 can be constructed by the following method. (FIG. 3) is a main part flow chart for explaining the operation when the judgment unit 3 is realized by another method.
It is. Here, the processing from step 30 to step 35 is the same as the processing from step 20 to step 25 in FIG. Also, it is assumed that the sound flag is turned on when it is determined that the sound is a sound in the previous F-frame determination, and that the music flag is turned on when it is determined that the music is the music.

[0022] First, the previous determination result in step 37 if silence counter is greater than a threshold L _T is determined whether a speech in step 36, if voice Step 38
Then, the determination flag is turned on to determine that the voice is sound. If the previous determination result is not a voice, only the voice flag is turned on, and the determination flag retains the previous state. If the condition is not satisfied in step 36, it compares the voice counter value and the threshold value H _T in step 40. If this condition is satisfied, it is determined in step 41 whether or not the previous determination result is music. If it is music, the determination flag is turned off and the music is determined to be music. If the previous determination result is not music, only the music flag is turned on in step 43, and the content of the determination flag is held as it is. If the condition of step 40 is not satisfied, both the voice flag and the music flag are turned off, and the contents of the determination flag are held as they are. By these processes, the content of the determination flag is changed only when the determination result for each F frame is continuously voice or music.

As described above, according to this embodiment, the sound / non-speech judging unit 2 determines whether the frame is sound or non-sound from the power value for each frame obtained by the power calculating unit 1. Then, the judgment unit 3 judges whether the sound is voice or music based on the frequency of occurrence of silence in the sound for each F frame based on the frequency of occurrence of silence in the sound. If neither can be judged, the previous F frame is used. -The system is configured to hold the results of the audio and music, so that the voice and music can be determined with high accuracy for the continuously input signal. Since the determination result is fixed to the dominant side, it is possible to prevent a phenomenon such as so-called chattering in which the determination of voice music changes every short time.

The determination unit 3 changes the determination result when the determination result for each F frame is determined to be voice or music twice consecutively, and otherwise retains the previous determination result. By doing so, the accuracy of the voice / music determination is further increased, and the determination can be switched smoothly.

[0025]

As described above, according to the present invention, a power calculation unit for calculating power for each frame, and a power value for each frame are compared with a preset power threshold. A sound / silence determining unit that determines whether the sound is sound or not, and determines whether the sound is music or not by comparing the appearance ratio of sound / silence for each of a plurality of frames with each threshold. In the case where the judgment is uncertain, by providing the judgment unit which holds the judgment results of the previous plurality of frames, it is possible to judge the sound and the music continuously with a short delay, and
It is possible to provide a voice / music discriminating apparatus that can output a stable judgment result even when a part that is difficult to judge is partially present.

The determination unit changes the determination result only when it is determined that the voice or music is successively determined in each of a plurality of frames, and otherwise, retains the determination result at the time of the previous determination as it is. By providing the audio / music discriminating apparatus, it is possible to further improve the discrimination accuracy and to smoothly perform the discrimination switching.

[Brief description of the drawings]

FIG. 1 is a block diagram of a voice / music discriminating apparatus according to an embodiment of the present invention;

FIG. 2 is a main part flowchart for explaining the operation of a determination unit of the embodiment;

FIG. 3 is a main part flowchart for explaining another configuration of the determination unit of the embodiment.

[Explanation of symbols]

 Reference Signs List 1 power calculation unit 2 sound / non-speech judgment unit 3 judgment unit

──────────────────────────────────────────────────続 き Continuation of the front page (56) References JP-A-49-130607 (JP, A) JP-A-58-208796 (JP, A) JP-A-61-94098 (JP, A) JP-A-2- 267599 (JP, A) JP-A-5-113797 (JP, A) JP-A-4-359298 (JP, A) (58) Fields investigated (Int. Cl. ⁶ , DB name) G10L 3/00 G10L 3 / 00 513 G10L 3/00 531 G10L 9/00

Claims (4)

(57) [Claims] 1. A power calculation unit (1), a sound / no-sound determination unit
(2) A voice / music discriminating apparatus including a judging unit (3).
The power calculation unit (1) converts the input signal into a frame
And the sound power of each frame is calculated, and the sound / non-speech determination unit (2) calculates the sound power of the power calculation unit (1).
If the sound power is greater than the sound power threshold,
If the interval is small, the section is determined to be a silent section, and the determination section (3) performs first and second determinations for each of a plurality of frames.
And do not apply to the first and second judgments
In this case, the previous determination is the current determination, and the first determination is that the number of frames determined to be a silent section is equal to the audio threshold.
If greater than the value, a plurality of frames is determined that speech, the second determination, the number of frames is determined that sound period music threshold
If it is larger than the value, multiple frames are determined to be music
Voice music discrimination device. 2. A power calculation unit (1), a sound / non-speech determination unit.
(2) A voice / music discriminating apparatus including a judging unit (3).
The power calculation unit (1) converts the input signal into a frame
And the sound power of each frame is calculated, and the sound / non-speech determination unit (2) calculates the sound power of the power calculation unit (1).
If the sound power is greater than the first sound power threshold.
Sound section, silent section when smaller than the second sound power threshold
And the determination unit (3) performs the first and second determinations for each of a plurality of frames.
And do not apply to the first and second judgments
In this case, the previous determination is the current determination, and the first determination is that the number of frames determined to be a silent section is equal to the audio threshold.
If greater than the value, a plurality of frames is determined that speech, the second determination, the number of frames is determined that sound period music threshold
If it is larger than the value, multiple frames are determined to be music
Voice music discrimination device. 3. A power calculation section (1), a sound / non-sound determination section.
(2) A voice / music discriminating apparatus including a judging unit (3).
The power calculation unit (1) converts the input signal into a frame
And the sound power of each frame is calculated, and the sound / non-speech determination unit (2) calculates the sound power of the power calculation unit (1).
If the sound power is greater than the sound power threshold,
If the interval is small, it is determined to be a silent section, and the determination unit (3) determines that the first plurality of frames and the second plurality
The first and second determinations are made for each frame, and the first and second determinations are made.
If the second judgment does not apply, the previous judgment is
The times of the determination, the first determination, a silent interval and determine the first plurality of frames
Number of frames is greater than the audio threshold and the second
The number of frames determined to be silent in the
If the value is larger than the value, the second multiple frames are determined to be audio
However , the second determination is made as a voiced section in the first plurality of frames.
Number of frames is greater than the music threshold and the second
The number of frames determined to be sounded in the
If the value is larger than the value, the second multiple frames are determined to be music
Voice and music discriminating device. 4. A power calculation section (1), a sound / non-sound determination section.
(2) A voice / music discriminating apparatus including a judging unit (3).
The power calculation unit (1) converts the input signal into a frame
And the sound power of each frame is calculated, and the sound / non-speech determination unit (2) calculates the sound power of the power calculation unit (1).
If the sound power is greater than the first sound power threshold.
Sound section, silent section when smaller than the second sound power threshold
The determination unit (3) determines that the first plurality of frames and the second plurality
The first and second determinations are made for each frame, and the first and second determinations are made.
If the second judgment does not apply, the previous judgment is
The times of the determination, the first determination, a silent interval and determine the first plurality of frames
Number of frames is greater than the audio threshold and the second
The number of frames determined to be silent in the
If the value is larger than the value, the second multiple frames are determined to be audio
However , the second determination is made as a voiced section in the first plurality of frames.
Boss was the number of frames is greater than the music threshold, the second plurality off
The number of frames determined to be sounded in the
If the value is larger than the value, the second multiple frames are determined to be music
Voice and music discriminating device.