JP2835483B2 - Voice discrimination device and sound reproduction device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a speech discriminating apparatus which is used as a pre-processing device for a video apparatus, a sound apparatus, or the like, and which automatically discriminates whether or not a continuously input sound signal is a sound. And a sound reproducing device using a voice discriminating device.

[0002]

2. Description of the Related Art In recent years, a function called "surround" for creating a sound effect has been added to a stereo device or a television receiver (hereinafter, referred to as a television). While these functions are very effective for sources such as music, they are less clear for sources mainly composed of audio such as news programs. Therefore, if it is possible to automatically determine whether the source is mainly of voice or not, the sound field and the frequency characteristics can be optimally controlled according to the result.

[0003] A conventional voice discriminating device utilizes that an input signal is a stereo signal. That is, in the case of a source such as music, a left channel (hereinafter, referred to as an L channel) and a right channel ( hereinafter, referred to as an R channel)
Are independent of each other, and the correlation between both channels is low. Conversely, in the case of a sound-based source such as a news program, the source is localized at the center, and the left signal (hereinafter, referred to as L signal) and the right signal (hereinafter, referred to as R signal) are almost the same signal. The correlation between both channels is high. Therefore, the difference between the amplitudes of the L signal and the R signal is calculated, and when the difference is small, it is determined that the signal is an audio signal, and when the difference is large, the signal is determined to be a signal other than voice. Further, the correlation value between the L signal and the R signal is calculated, and if the correlation value is large, the signal may be an audio signal, and if the correlation value is small, the signal may be a signal other than audio.

[0004]

BRIEF Problem to be Solved] In the conventional voice discriminating apparatus, that is effective for the source of the scan Te Leo, can not be determined for the L signal and the source of the mono no difference in R signal There was a problem.

SUMMARY OF THE INVENTION The present invention solves the above-mentioned problems. An audio discriminating apparatus capable of accurately discriminating whether a signal is a monaural signal or a stereo signal is provided. It is an object of the present invention to provide a sound reproducing apparatus that can automatically control sound characteristics in accordance with the sound reproduction.

[0006]

According to a first aspect of the present invention, there is provided a power calculating unit for calculating the acoustic power of an acoustic signal for each frame of a predetermined time, and a threshold value for setting the computed acoustic power value in advance. A sound / silence determining unit that determines the sound / non-sound of the frame by comparing with a zero-crossing calculating unit that calculates the number of zero-crossings of the waveform of the acoustic signal for each frame, and calculates the calculated number of zero-crossings in advance. A consonant determining unit that determines a consonant of the frame by comparing the set threshold value with a set threshold value , and detects a maximum value and a minimum value of power values in a plurality of continuous predetermined frame periods, and calculates a difference value between the detected power values. The sex determination unit, the presence ratio of the frames determined to be silent in the plurality of frames, the presence ratio of the frames determined to have high consonantness, and the difference value are preset in advance. And the voice if all than the value large
Is determined and the voice is not determined.
Existence ratio of frames judged to be silent in the frame
And the difference value are respectively set to the threshold value in advance.
Multiple if smaller than the set threshold
The sound signal in a frame is determined to be non-voice, and in other cases, the sound signal in a plurality of frames is determined to be undefined.
Constant, and a voice discriminating apparatus and a sound determination unit for outputting a determination result for each of a plurality frames, also invention, line discrimination of speech non-speech to input acoustic signals according to claim 2
And the sound signal and the sound signal.
The voice non-voice discrimination result for each predetermined time of the voice discrimination device is
Input, and the frequency of the sound signal is
The frequency at which the wave number characteristic is gradually changed to the optimal frequency characteristic
This is a sound reproduction device including a characteristic control unit.

[0007]

According to the first aspect of the present invention, the power calculation section calculates the signal power of the frame section of the audio signal, determines whether the section is sound or no sound based on the calculated power level, and performs zero-crossing. The number calculation unit calculates the number of zero crossings in the frame section of the audio signal, and the consonant determination unit determines the consonantness of the section from the magnitude of the calculated number of zero crossings. The continuity determining unit calculates a difference value between a maximum value and a minimum value of power in a plurality of continuous frame sections. The sound discriminating unit determines that the sound signal in the plurality of frame sections is not sound when the existence rate of silence frames, the existence rate of consonant frames, and the power difference value are larger than the threshold values set respectively. And the sound
And the existence ratio of silence frames
And the power difference values are smaller than the respective threshold values.
If it is smaller than the set threshold, it is judged as non-voice
Otherwise, it is determined to be undefined.

Further, in the invention according to claim 2, the voice discriminating section determines whether or not the acoustic signal is a voice, and determines a frequency characteristic.
The sex control unit, based on the determination result,
Frequency characteristics of the sound signal
And switch the output.

[0009]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS (Embodiment 1) An embodiment of the voice discriminating apparatus of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram showing the configuration of this embodiment. In the figure, 1 is a power calculation unit for calculating the power of an input signal, 2 is a zero-crossing calculation unit for calculating the number of zero-crossings of the waveform for each frame, and 3 is a unit that compares the calculated power with a threshold value. A voiced / silent determining unit that determines whether the input signal of the frame is voiced or silent, 4 is a consonant determining unit that determines the presence or absence of consonantness of the frame based on the number of zero crossings for each frame, 5 stationarity determining unit for determining the stationarity by the difference value between the maximum value and the minimum value of the power of every predetermined plurality of frames, a silence determination frame number ratio of the plurality frames 6, the maximum power among a plurality of frames And the difference between the minimum power and the ratio of the number of frames where the number of zero crossings in a plurality of frames is a certain number or more,
An audio determination unit that determines whether the audio is audio or non-audio for each of a plurality of frames.

The interrelationship and operation of the above components will be described. Here, the input signal is a signal of a device such as an audio device and a television, and is a stereo signal. The input L signal and R signal of the stereo signal are mixed, and (L
+ R) as a signal. The power calculation unit 1 calculates, for each frame at a fixed time interval, a cumulative value or an average value of amplitudes in the section as a power value in the frame. The zero-crossing calculation unit 2 calculates the number of times the input waveform crosses the zero amplitude value for each frame as the number of zero crossings Z0. In the case of speech, the number of zero crossings shows a large value especially for unvoiced fricative consonants. The consonant determining unit 4 determines that the consonant is high if the number of zero crossings Z0 of the frame obtained by the zero crossing calculating unit 2 satisfies Z0> Zt. Here, Zt is a preset threshold value for judging consonantness. According to the experimental results, when the sampling frequency is 10 kHz and the frame length is 20 milliseconds, about 40 times is a reasonable value. . The number of frames determined to have high consonantness is accumulated in units of a fixed number of frames. This accumulated value is assumed to be NZ.

The sound / non-speech determining unit 3 uses the power value obtained by the power calculating unit 1 to determine whether there is sound or no sound for each frame. Here, assuming that the power value of the current frame is P and the threshold value of the sound / silence determination is Pt, when P <Pt is satisfied, it is determined that there is no sound, and the number of frames determined to be silent is determined in units of a fixed plurality of frames. Cumulative. The accumulated value of the number of frames is defined as Np.

Here, the threshold value Pt is a preset value, but may be adaptively determined according to a change in the input level. The above process is a process of one frame unit.

The following processing is performed with a plurality of F frames as one unit. Here, the processing interval F is a unit in which the features of the voice can be confirmed at a minimum, and in the case of a voice uttered continuously, a value including an average of a few syllables (for example, 1 second to 2 seconds) Seconds).
The larger the value F is, the more accurately the voice-likeness can be detected. However, the time required for the determination becomes longer, and thus the value F is determined by a trade-off between the two.

From the cumulative value NZ of the number of frames determined to have high consonantness in the F-frame section and the cumulative value NP of the number of frames determined to be silent, the existence ratio of frames having high consonantity in the F-frame section is NZ /. The existence ratio of the silent section in the F and F frame sections is given as NP / F.

Further, the continuity determining unit 5 detects the maximum value and the minimum value of the power during each F frame, and calculates the difference value Pd. Since the continuously uttered voice is a repetition of a vowel, a consonant, and a silent part, the power change, that is, the value of Pd naturally increases at a certain time interval (in this case, F frame). Therefore, the value of the value Pd is used as a target for determining the likelihood of speech.

The voice determining unit 6 uses the Nz, Np, and Nd obtained by the voiced / silent determining unit 3, the continuity determining unit 4, and the consonant calculating unit 5, respectively, to determine the existence ratio of the silent section and the consonantness. Is determined to be a voice when the conditions of the existence ratio of frames having a high frame rate and the power difference value, that is, all of the following determination formulas are satisfied.

A <(NZ / F) <b (Np / F)> c Pd> Pdtv where a, b, c, and Pdtv are threshold values for each parameter for sound / silence determination. Determine the optimal value. a and b are the lower and upper thresholds of the existence ratio of frames with high consonance, c is the threshold of the existence ratio of silence periods, and Pdtv is a threshold for measuring the degree of power change. According to the above processing, when the silence section and the consonant section exist in the F frame at a certain value or more and the power change is large, it is determined that the source is likely to be speech, and is determined to be speech. On the other hand, non-speech determination is performed when there is almost no silent section and the power change is small (constant), especially when non-speech is limited to music.
That is, only when the condition of (Np / F) <dPd <; Pdtu is satisfied, it is determined that the sound is non-voice (music). Here, d is a threshold value of the existence ratio of a silent section for non-speech determination, and Pdtu is a threshold value for measuring the degree of power change for non-speech determination.
D <c and Pdtu <Pdtv.

If neither the speech nor the non-speech determination condition is satisfied, the result is indeterminate because neither can be determined. It is possible to prevent an erroneous determination by determining that this is indefinite, and to prevent a phenomenon in which the determination of voice or non-voice is switched in a short time by retaining the previous determination result in the case of indeterminate. it can. These series of determination results are continuously output from the voice determination unit 6 at the F frame period.

As described above, according to the present embodiment, the sound signal is determined to be a speech when the consonant existence ratio, the silence existence ratio, and the maximum and minimum difference values of the sound power are larger than the respective predetermined values. By doing so, it can be determined whether or not the audio signal is audio regardless of whether the audio signal is monaural or stereo. If the audio signal is not audio, the continuity of the signal and the difference between the maximum and the minimum are smaller than the audio, so that the audio signal is like music. It is possible to determine with high accuracy whether the voice is a non-voice . Also voice
If it is not possible to determine whether it is non-speech
Causes extreme switching due to erroneous judgment between speech and non-speech.
Can be prevented.

(Embodiment 2) Hereinafter, an embodiment of a sound reproducing apparatus according to the present invention according to claim 2 will be described with reference to the drawings.

FIG. 2 is a block diagram showing the configuration of this embodiment. In the figure, reference numeral 7 denotes an audio / music discriminating unit, which outputs a judgment result as to whether the section is sound or music at regular intervals. Reference numeral 8 denotes a frequency characteristic control unit that gradually switches to a frequency characteristic suitable for voice or music based on the determination result of the audio / music determination unit 7. FIG. 3 shows an example of a frequency characteristic diagram in which the frequency characteristic control section 8 switches.

The operation of the above configuration will be described. First, the audio / music discriminating unit 7 receives the (L + R) signal, determines whether the signal is audio, music, or indefinite at regular intervals (F frame intervals), and outputs the result to the frequency characteristic control unit 8. The operation of the audio / music discriminating section 7 is the same as the operation of the audio discriminating apparatus in the first embodiment, and a description thereof will be omitted. Also, non-voice is considered music here. The frequency characteristic control unit 8 is set in advance, for example, as shown in FIG.
There are 10 frequency characteristics as shown in
If the input signal is an audio source, control is performed so that the frequency characteristic finally becomes one, and if the input signal is a music source, control is performed so that the frequency characteristic becomes ten.

It is assumed that the frequency characteristic is set to 5 as an initial state. When the judgment result of the sound is received from the sound / music judgment section 7, the characteristic is changed to 4 in order to approach the characteristic 1 of the one-step sound. When the judgment result of music is received, on the contrary, the characteristic is approximated by one step to the characteristic of 10 and is changed to the characteristic of 6. If the determination result is indeterminate, the current state 5 is maintained. By repeating this operation based on the sound / music discrimination result sent for each F frame, for example, if the judgment result of sound continues, the characteristic gradually approaches a characteristic suitable for sound reproduction, and finally a characteristic of 1 And the state is fixed until the next judgment result of music is received.

As described above, according to the present embodiment, the audio / music discriminating section 7 for discriminating whether the source is voice or music, and the frequency gradually approaching the frequency characteristic suitable for the source based on the discrimination result. By providing the characteristic control unit 8, the frequency characteristics of the device can be automatically changed to the frequency characteristics suitable for the input source, and a sound reproduction device that is easy to hear can be realized. In addition, the characteristics are not switched all at once to the optimum characteristics for voice and music, but are switched stepwise so that the user does not feel uncomfortable due to the change in the frequency characteristics. Ma
Also, if it is not possible to judge both voice and music, judge it as indeterminate
And by controlling not to change the characteristics,
Phenomenon that characteristics are switched in a short time due to erroneous judgment
And output the playback sound suitable for the source without discomfort
be able to.

[0026]

As is apparent from the above description, the invention according to claim 1 is a power calculating unit for calculating the acoustic power of an acoustic signal for each frame of a fixed time, and the calculated acoustic power value is set in advance. A voiced / silent determination unit that determines voiced / voicelessness of the frame by comparing with the threshold value, a zero-crossing calculation unit that calculates the number of zero-crossings of the waveform of the acoustic signal for each frame, A consonant determining unit that determines the consonantness of the frame by comparing the number of crossings with a preset threshold value, and detects the maximum value and the minimum value of the power value in a plurality of continuous predetermined frame periods, and calculates the difference value. The calculated continuity determination unit, the existence ratio of the frames determined to be silent in the plurality of frames, the existence ratio of the frames determined to have high consonantness, and the difference value are each previously determined. When all of the sound signals are larger than the set threshold value, the sound signal in the plurality of frames is determined to be sound, and the sound signal in the plurality of frames is not determined to be sound. If the value is smaller than the threshold value set in advance to be smaller than the threshold value, the sound signal in the plurality of frame sections is determined to be non-speech, and otherwise, the sound signal in the plurality of frames is undefined. By providing a sound determination unit that makes a determination and outputs a determination result for each of a plurality of frames, it is possible to accurately determine whether a sound signal is sound or non-speech regardless of whether the sound signal is monaural or stereo . In addition, voice and non-voice
If you are not sure, you can determine
The judgment that voice / non-voice is suddenly switched by the judgment
Can be prevented.

According to a second aspect of the present invention, there is provided a speech discriminating apparatus for discriminating speech non-speech by inputting a speech signal, and a speech non-speech discrimination result of the speech signal and the speech discriminating apparatus at predetermined time intervals. And a frequency characteristic control unit for outputting a frequency characteristic in a stepwise manner in accordance with a result of the voice non-voice discrimination and changing the frequency characteristic to a characteristic optimum for the audio signal.
With this, regardless of whether the sound signal is monaural or stereo , the sound signal can be automatically reproduced with the frequency characteristic corresponding to whether it is sound or not , and in the case of indeterminate judgment
Erroneous judgment by maintaining the previous characteristics
Sudden change in frequency characteristics due to
Prevents the phenomenon that the characteristics of the non-voice side are unstablely changed, and
It is possible to provide a sound reproducing device that is always easy to hear.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of an embodiment of a voice discrimination device of the present invention.

FIG. 2 is a block diagram showing a configuration of an embodiment of a sound reproducing apparatus according to the present invention.

FIG. 3 is a frequency characteristic diagram of an embodiment in which a frequency characteristic control unit in the sound reproducing device of the present invention switches stepwise.

[Description of Signs] 1 Power calculation unit 2 Zero-crossing calculation unit 3 Voiced / silence determination unit 4 Consonant determination unit 5 Stationaryness determination unit 6 Voice determination unit 7 Voice / music determination unit 8 Frequency characteristic control unit

──────────────────────────────────────────────────続 き Continued on the front page (58) Field surveyed (Int. Cl. ⁶ , DB name) G10L 3/00 531 G10L 3/00 513 G10L 9/12 301

Claims (2)

(57) [Claims] 1. A sound of an acoustic signal for each frame of a predetermined time.
A power calculation unit for calculating the sound power, and the calculated sound power.
The power value is compared with a preset threshold value and
A sound / silence determining section for determining sound / silence of a frame;
Calculate the number of zero crossings of the sound signal waveform for each frame.
The zero-crossing calculation unit and the calculated number of zero-crossings are set in advance.
Judge the consonantness of the frame by comparing it with the threshold
The consonant determination unit and a plurality of continuous
The maximum and minimum power values are detected and the difference is calculated.
And a continuity determining unit that emits no sound in the plurality of frames.
And the consonant is determined to be high.
The defined frame existence ratio and the difference value are respectively
If all values are greater than the preset threshold
In this case, the audio signal in the
Is not determined to be audio,
The existence ratio of frames determined to be silent in the
Each of the difference values is equal to or less than the threshold value in advance.
If it is smaller than the threshold set in
The sound signal in the frame section is determined to be non-voice, and the other
Sound that is judged to be constant and the judgment result is output for each frame
A voice discriminating device comprising a voice determining unit. 2. An audio signal is inputted to determine whether a voice is a non-voice.
2. The voice discriminating apparatus according to claim 1, wherein said audio signal is
The voice non-voice discrimination result of the voice discrimination device every predetermined time and
Of the sound signal according to the voice / non-voice discrimination result.
Frequency that gradually changes the frequency characteristics to the optimal frequency characteristics
An audio reproducing device comprising a numerical characteristic control unit.