JPH05113797A - Voice/musical-sound discriminating circuit - Google Patents

Abstract

PURPOSE:To efficiently discriminate with high quality both a voice signal and a musical-sound signal from a digitalized input signal. CONSTITUTION:The self-correlation coefficient R(t) of an input signal is obtained by a self-correlation coefficient calculating means 2. A first threshold value judging integrating means 10 integrates the number of occurrence of the values R(t) greater than a positive threshold value PK1. A second threshold value judging integrating means 20 integrates the number of occurrence of the values of R(t) less than a negative threshold value PK2. An integration frequency judging means 30 judges whether or not the threshold values of the means 10, 20 exceed respective threshold values. A counting discriminating means 40 counts the continuity within a given interval with regard to the result judged by the means 30 and carries out the discrimination between a voice signal and a musical-sound signal.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a voice / musical tone discrimination circuit for discriminating between a voice signal spoken by a human and a musical tone signal generated by a musical instrument.

[0002]

2. Description of the Related Art Conventionally, as a technique in such a field,
For example, some documents were described in the following documents. References: 1987 IEICE Information and Systems Division National Convention Proceedings 380 Hosoda et al. "A Study on Data Identification Method for DCME Overload Protection" P. 2-111 Conventionally, as described in the above document, in the field of international communication using satellites, etc., DSI (data discrimination
algorithm) technology and ADPCM (adaptive pulse-code)
A digital line multiplexer (hereinafter referred to as "DCME"), which is a combination of modulation technologies, has been receiving attention. This DC
In the ME, when the transmission channel is overloaded, a method of reducing the allocated bit (usually 4 bits) of the transmission channel by 1 bit and eliminating the lockout of the call is used. If the number of channels for transmitting a modem signal composed of voice band data) is reduced by 1 bit, the transmission characteristics deteriorate. Therefore, it is necessary to identify whether the input signal of ADPCM is voice band data. Therefore, in the technique of the above document, the data identification algorithm for DCME overload protection is used to identify whether the input signal is a voice signal or voice band data.

[0003]

However, if a voice / musical tone discrimination circuit for discriminating between a voice and a musical tone is constructed by using the conventional method for discriminating between a voice signal and a modem signal (voiceband data), the following is obtained. There were challenges.

Here, a musical tone means a sound that can be separated from a voice signal spoken by humans. The audio signal is
It is a means of communication for realizing conversation, and it is a sound that can ensure a certain level of quality. Speech is generated from the human vocal tract and is given spectral characteristics mainly by the formants. Further, the musical tone is a musical tone that gives a rich musical feeling, and the musical tone has a characteristic that vibrations are periodic and harmonic frequencies form a positive ratio. As an example of these, FIG. 2 is a time-series waveform diagram of an audio signal, and FIG.
The time-series waveform charts of the musical tone signals are shown in each. The tone signal of FIG. 3 is a waveform diagram by the orchestra. 2 and 3
The waveform charts drawn under the respective waveform charts are the enlarged views of the above figures.

When distinguishing such a voice signal and a tone signal, since the characteristics of the two signals are too different, the voice signal and the tone signal are distinguished from each other, and the voice signal has a band quality corresponding to the voice signal. It has been difficult to provide a high-quality and efficient voice / musical sound discrimination circuit by ensuring the intelligibility and improving the intelligibility and handling the musical sound signal in a wide band corresponding to the musical sound signal.

The present invention has the problems that the above-mentioned prior art has, and distinguishes a voice signal from a musical tone signal, improves the intelligibility and the intelligibility of the voice signal, and also provides a high-quality and efficient musical tone signal. The present invention provides a speech / tone identification circuit that solves the problem of difficulty in identifying.

[0007]

In order to solve the above problems, the present invention provides a voice / tone identification circuit for identifying a voice signal and a tone signal from a digitized input signal. Autocorrelation coefficient calculation means for obtaining the number of relations, and first threshold value judgment integration means for integrating the number of times that the autocorrelation coefficient has a value larger than a positive threshold value,
Whether or not the integrated value of the second threshold value judgment integrating means for integrating the number of times that the autocorrelation coefficient is smaller than the negative threshold value and the integrated value of the first and second threshold value judgment integrating means exceed the threshold value And a counting and identifying means for identifying a voice signal or a musical tone signal by counting the continuity of a certain section with respect to the determination result of the integrating number determining means.

[0008]

According to the present invention, since the voice / tone discrimination circuit is constructed as described above, when the digitized signal is input, the autocorrelation coefficient calculating means obtains the autocorrelation coefficient and the 1, sent to the second threshold judgment integrating means. The first threshold value judgment integrating means judges whether or not the autocorrelation coefficient is larger than a positive threshold value, and counts the number of times a value larger than the positive threshold value is generated. Similarly, the second threshold value judgment integrating means judges whether or not the autocorrelation coefficient is smaller than the negative threshold value, and counts the number of times the small value occurs.

The integrated number determination means determines whether or not the two integrated values of the first and second threshold value determining and integrating means respectively exceed the threshold value, or whether the sum of the two integrated values exceeds the threshold value. Etc. and sends the result of the judgment to the counting and identifying means. The count identifying means counts the continuity of the number of times generated by the integrated number determining means, and distinguishes between the voice signal and the musical tone signal. Therefore, the above problem can be solved.

[0010]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS FIG. 1 shows a voice / audio system according to a first embodiment of the present invention.
It is a functional block diagram of a musical sound identification circuit. This voice / musical sound discrimination circuit calculates the power of the digitized input signal Xi to determine whether there is voice or no voice.
And an autocorrelation coefficient calculating means 2 for calculating an autocorrelation coefficient R (t) from the input signal Xi based on the presence judgment of the power calculating means 1. On the output side of the autocorrelation coefficient computing means 2, first and second threshold value judgment integrating means 10, 2 are provided.
0 is connected.

The first threshold judgment integrating means 10 has a function of integrating the number of times that the autocorrelation coefficient R (t) is larger than the positive threshold value PK1, and the autocorrelation coefficient R (t) is calculated. The threshold value determining means 11 that outputs when the autocorrelation coefficient R (t) is larger than the positive threshold value PK1 and the number of outputs of the threshold value determining means 11 are integrated (counted) and the integrated value M1 is output. And the determination number integration means 12 for performing the determination.
Similarly, the second threshold judgment integrating means 20 has a function of integrating the number of times the autocorrelation coefficient R (t) is smaller than the negative threshold PK2, and the autocorrelation coefficient R (t). Is compared with a negative threshold value PK2 to output when the autocorrelation coefficient R (t) is small, and the threshold value determining means 2
The determination number integration means 22 that integrates the number of times of output of 1 and outputs the integrated value M2.

First and second threshold value judgment integrating means 10, 20
The cumulative number judging means 30 is connected to the output side of, and the count identifying means 40 is connected to the output side thereof. The integrated number determination means 30 determines whether or not the integrated values M1 and M2 exceed the thresholds K1 and K2, and outputs the determination result of the number of times when the conditions of M1> K1 and M2 <K2 are satisfied. have.

The count identifying means 40 is a cumulative number judging means 3
It has a function to count the continuity of the number of occurrences of 0,
Two count-up conditions (CNT1, CNT2) are connected so that they are input to other resets R1, R2.
Two counters 41 and 42, an inverter 43 that inverts the output of the integration number determination means 30, and a carrier CR1 of the counter 41 to reset the carrier C of the counter 42.
The flip-flop (hereinafter referred to as FF) 44 is reset by R2 and outputs the tone signal Y1 or the audio signal Y2 from the positive-phase output terminal and the negative-phase output terminal.

Next, the operation of FIG. 1 will be described with reference to FIGS. 4 and 5. FIG. 4 is an operation flowchart of FIG. 1, and S1 to S17 represent processing steps. Also, FIG.
2 is a diagram showing the autocorrelation coefficient R (t) of FIG. 1, in which the output is normalized to a maximum of 1. FIG.

In FIG. 4, when the operation of the voice / tone discrimination circuit of FIG. 1 starts, the digitized input signal Xi
Are stored in a memory (not shown) in block units (frame units) in step S1. The block-by-block input signal Xi stored in the memory is input to the power calculation means 1 and the autocorrelation coefficient calculation means 2. The power calculation means 1 calculates the power of the input signal Xi in step S2, and the signal power is the threshold Pt in step S3.
It is determined whether the signal power is larger than h, and the signal power is the threshold Pth.
When it is larger, it is considered as voiced and an instruction for calculation is issued to the autocorrelation coefficient calculation means 2. Other than that, it is considered to be silent, and the operation is ended in this block section without distinguishing between the voice signal and the tone signal.

In the autocorrelation coefficient calculation means 2, step S
4, the autocorrelation coefficient R (t) is calculated by the following equation.
To calculate.

[0017]

[Equation 1]

The calculated autocorrelation coefficient R (t) is shown in FIG.
The waveform is as shown in (1) and is sent to the first and second threshold value judgment integrating means 10, 20.

In step S5 of FIG. 4, the integrated values M1 and M2 of the determination number integration means 12 and 22 are initialized to zero. In step S7 through step S6, the threshold value determination means 11 sets the threshold value PK1 for which the autocorrelation coefficient R (t) is positive.
Is compared with the positive threshold PK1 and the autocorrelation coefficient R
When (t) is large, it is output to the judgment number integrating means 12. The determination count integration means 12 counts up in step S8. On the other hand, in the threshold value judging means 21, the autocorrelation coefficient R (t) and the negative threshold value P in step S9.
K2 and the autocorrelation coefficient R (t) is a negative threshold P
When it is smaller than K2, it is output to the judgment number integrating means 22. The determination count integration means 22 counts up in step S10. These judgment counts are integrated in step S
Repeated through 6 for the number of signals N in the block.

Integrated value M of the judgment number integrating means 12, 22
1 and M2 are sent to the integrated number determination means 30. In the cumulative number determination means 30, at the end of one block, step S1
At 1, it is determined whether or not the conditions of M1> K1 and M2 <K2 are satisfied, and the determination result is given to the counter 41 in the count identifying means 40 and also inverted by the inverter 43 and given to the counter 42. When the conditions of M1> K1 and M2 <K2 are satisfied in step S11, step S12
Then, the counter 41 counts up and is reset R1
Causes the counter 42 to be reset to zero. Conversely, when the conditions of M1> K1 and M2 <K2 are not satisfied in step S11, the counter 42 counts up in step S13, and the counter 4 is reset by the reset R2.
1 is reset to 0.

The counters 41 and 42 output the carriers CR1 and CR2 as overflows when they have counted a predetermined number of times. This counter 41, 42
Then, the continuity of the number of times generated by the integrated number determination means 30 which is usually about 2 to 4 is counted. That is, the counter 41
Counts up when the determination by the integration number determination means 30 is true, outputs the carrier CR1 due to the count overflow, sets the FF 44, and determines the tone signal Y1 by the positive phase output of the FF 44 in step S14. Similarly, the counter 42 counts up when the determination by the number-of-accumulations determination means 30 is false, resets the FF 44 due to a count overflow, and determines in step S16 and S17 that the audio signal Y2 is the reverse phase output of the FF 44. ..

When the identification process within one block is completed in this way, the identification process for the next block is performed. As described above, in the present embodiment, the number of times the value of the autocorrelation coefficient R (t) is larger than the positive threshold value PK1 is counted, and the number of times the value is smaller than the negative threshold value PK2 is counted. Since the integration number determination means 30 determines whether or not the positive and negative count values exceed the threshold value, the musical tone signal Y1 and the voice signal Y2 can be clearly detected. Moreover, the tone signal Y1 and the voice signal Y2 can be identified in a short time, for example, 3 to 4 block time. Furthermore, since the amount of calculation is small, the circuit scale can be reduced.

FIG. 6 is a functional block diagram of a voice / tone identification circuit showing a second embodiment of the present invention. Elements common to those in FIG. 1 are designated by common reference numerals.

In this voice / musical sound discrimination circuit, an integrated number judging means 30A having a different structure is provided in place of the integrated number judging means 30 of FIG. The cumulative number judging means 30A judges whether or not the sum of the cumulative values M1 and M2 of the positive and negative judging frequency cumulative means 12 and 22 exceeds a threshold value K, and when M1 + M2> K, the output is true. So that the same action and effect as those of the first embodiment can be obtained.

The present invention is not limited to the above embodiment,
Various modifications are possible. Examples of such modifications include the following. (A) The count identifying means 40 of FIGS. 1 and 6 may be configured using other circuit configurations. (B) The circuits shown in FIGS. 1 and 6 may be composed of individual circuits using integrated circuits or the like, or may be digital signal
You may perform by program control using a processor (DSP) etc. When executed by program control, the circuit scale can be reduced as compared with an individual circuit.

[0026]

As described in detail above, according to the present invention, the autocorrelation coefficient of an input signal is obtained, and the number of times that the autocorrelation coefficient has a value larger than a positive threshold value is integrated and , Accumulate the number of times a value smaller than the negative threshold is generated,
It is determined whether or not the integrated value exceeds a threshold value, and a voice or a musical sound is identified by the determination result being continuous for a certain period. Therefore, by clearly distinguishing the audio signal from the musical sound signal, and by ensuring the quality of the band corresponding to the audio signal, the clarity and intelligibility are improved, and the musical sound signal corresponds to the musical sound signal. By handling a wide band, high-quality and efficient identification processing is possible.

Furthermore, since the identification process can be performed in a short time and the amount of calculation can be small, the circuit scale can be reduced. Therefore, the present invention can be applied to various signal processing devices that process voice signals and tone signals with digital signals.

[Brief description of drawings]

FIG. 1 is a functional block diagram of a voice / tone identification circuit according to a first embodiment of the present invention.

FIG. 2 is a time-series waveform diagram of an audio signal.

FIG. 3 is a time-series waveform chart of a musical tone signal.

FIG. 4 is an operation flowchart of FIG.

5 is a waveform chart showing the autocorrelation coefficient of FIG. 1. FIG.

FIG. 6 is a functional block of a voice / tone identification circuit showing a second embodiment of the present invention.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Power calculation means 2 Autocorrelation coefficient calculation means 10, 20 1st, 2nd threshold value judgment integration means 11, 21 Threshold value judgment means 12, 22 Determination frequency integration means 30, 30A Integration frequency determination means 40 Count identification means

[Procedure amendment]

[Submission date] December 26, 1991

[Procedure Amendment 1]

[Document name to be amended] Statement

[Name of item to be corrected] 0002

[Correction method] Change

[Correction content]

[0002]

2. Description of the Related Art Conventionally, as a technique in such a field,
For example, some documents were described in the following documents. References: 1987 IEICE Information and Systems Division National Convention Proceedings 380 Hosoda et al. "A Study on Data Identification Method for DCME Overload Protection" P. 2-111 Conventionally, as described in the above-mentioned document, in the field of international communication using satellites, etc., DSI ( Digital Speech Intep
olation) technology and ADPCM (adaptive diffe rential
A digital line multiplexer (hereinafter, referred to as DCME) that fuses pulse-code modulation) technology is drawing attention. In this DCME, when the transmission path channel is overloaded, a method of reducing the allocated bit (usually 4 bits) of the transmission path by 1 bit and eliminating the lockout of the call is used. (Voiceban
1 for the channel transmitting the modem signal composed of ddata)
If the number of bits is reduced, the transmission characteristics deteriorate. Therefore, A
It is necessary to identify whether the input signal of the DPCM is voice band data. Therefore, in the technique of the above document, DC
The ME overload protection data identification algorithm is used to identify whether the input signal is a voice signal or voice band data.

[Procedure Amendment 2]

[Document name to be amended] Statement

[Correction target item name] 0012

[Correction method] Change

[Correction content]

First and second threshold value judgment integrating means 10, 20
The cumulative number judging means 30 is connected to the output side of, and the count identifying means 40 is connected to the output side thereof. The integrated number determination means 30 determines whether or not the integrated values M1 and M2 exceed the thresholds K1 and K2, and outputs the determination result of the number of times when the conditions of M1> K1 and M2> K2 are satisfied. have.

[Procedure 3]

[Document name to be amended] Statement

[Correction target item name] 0020

[Correction method] Change

[Correction content]

Integrated value M of the judgment number integrating means 12, 22
1 and M2 are sent to the integrated number determination means 30. In the cumulative number determination means 30, at the end of one block, step S1
At 1, it is determined whether or not the conditions of M1> K1 and M2> K2 are satisfied, and the determination result is given to the counter 41 in the count identifying means 40 and inverted by the inverter 43 to be given to the counter 42. In step S11, M1> K1 and
When the condition of M2> K2 is satisfied, step S12
Then, the counter 41 counts up and is reset R1
Causes the counter 42 to be reset to zero. Conversely, when the conditions of M1> K1 and M2> K2 are not satisfied in step S11, the counter 42 counts up in step S13, and the counter 4 is reset by the reset R2.
1 is reset to 0.

[Procedure amendment 4]

[Document name to be corrected] Drawing

[Name of item to be corrected] Figure 1

[Correction method] Change

[Correction content]

[Figure 1]

[Procedure Amendment 5]

[Document name to be corrected] Drawing

[Name of item to be corrected] Fig. 4

[Correction method] Change

[Correction content]

[Figure 4]

Claims (1)

[Claims] 1. An autocorrelation coefficient calculating means for obtaining an autocorrelation coefficient of a digitized input signal, and a first threshold value judgment for integrating the number of times the autocorrelation coefficient has a value larger than a positive threshold value. Accumulating means, second threshold value judging and accumulating means for accumulating the number of times that the autocorrelation coefficient is smaller than a negative threshold value, and integrated values of the first and second threshold value judging and accumulating means exceed a threshold value. And a count identifying unit that identifies the audio signal or the musical tone signal by counting the continuity of a certain section with respect to the determination result of the integrating number determining unit. Voice / musical sound discrimination circuit.