JP3071476B2 - Voice pitch extraction 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 voice pitch extracting device, and more particularly to a voice pitch extracting device suitable for extracting a voice pitch by a voice recognition device.

[0002]

2. Description of the Related Art FIG. 6 is a block diagram showing a configuration of a conventional voice pitch extracting device. The audio signal 1 is converted into digital data 3 by the A / D converter 2. Digital data 3
FIG. 7 shows a speech waveform as an example. Speech is a signal in which vocal cord vibrations are emitted through the vocal tract, and has a pitch period, that is, the strongest amplitude in a vocal cord vibration period. Therefore, in order to obtain a pitch period from the digital data 3, the digital data 3 is input to the autocorrelation unit 21, and the first to 80th orders shown in FIGS. 8 and 9 (Kyoritsu Shuppan Co., Ltd. (See pages 53 to 59), and the order having the highest correlation among the autocorrelations is applied to the following equation 1 to obtain the pitch period signal 12.

[0003]

## EQU1 ## Pitch period = order of highest correlation × (1 / sampling frequency)

[0004]

However, in the above-described conventional voice pitch extracting apparatus, the autocorrelation unit 2
1 is required. Therefore, when trying to extract the maximum pitch period of 10 ms (100 Hz) (see the above-mentioned document) of a male by the autocorrelation, in an environment of a sampling frequency of 8 KHz, at least the first order is 10 × 10 ⁻³ (1/8000) = 80.
The following autocorrelation calculation is required. If the section for calculating the autocorrelation is 20 mS, the number of multiply-accumulations required for one autocorrelation is 20 × 10 ⁻³ / (1/8000) = 160 times, and the autonomous from the first to the 80th order is calculated. When the correlation is found, 160 × 80 = 1
2800 sum-of-products calculations are required. That is, in the conventional voice pitch extraction device, the autocorrelation unit 21 needs to perform many calculations within a fixed time, and it is necessary to incorporate a high-speed operation device such as a processor dedicated to signal processing as hardware. There is a problem that the device cost becomes extremely high.

The present invention has been made in view of such a problem, and it is possible to extract a voice pitch without using a correlator that requires many product-sum operations, thereby reducing the number of operations. An object of the present invention is to provide a voice pitch extraction device.

[0006]

According to the present invention, there is provided an audio pitch extracting apparatus for converting an audio signal into a digital audio signal.
/ D converter, a zero-cross detector that detects a change in the sign of the digital audio signal, a square wave generator that generates an audio square wave signal from the change in the sign, a digital VCO unit which outputs a square wave signal of corresponding frequency, compares the phases of the square wave signal and the audio square wave signal, and a digital phase comparator for outputting a phase lag signal and the phase advance signal, the phase The value obtained by integrating the delay signal and the phase advance signal is referred to as the digital VC.
An integrating section for outputting to the O section, and measuring an output of the integrating section;
Characterized in that it closed the terminal.

[0007]

In the present invention, a digital phase comparator, an integrator, and a digital VCO can form a PLL (Phase Locked Loop), and the digital voltage output from the integrator is an audio signal from the square wave generator. It changes following a square wave signal. Then, the frequency of the audio square wave signal at this time can be inversely calculated from the relationship between the digital voltage value in the digital VCO unit and the frequency of the square wave signal. Thereby, a voice pitch frequency is obtained.

[0008]

Next, embodiments of the present invention will be described with reference to the accompanying drawings.

FIG. 1 is a block diagram showing a configuration of a voice pitch extracting apparatus according to an embodiment of the present invention. Audio signal 1 is A
The digital data 3 is converted by the / D converter 2 into digital data 3, and the digital data 3 is sent to the zero-cross detector 4. The zero-cross detecting unit 4 detects a point where the digital data 3 changes from-to + and + to-. When the digital data 3 changes from-to +, 1 is set. To the square wave generator 6 as zero-cross data 5
Send to The square wave generator 6 generates an audio square wave signal 7 which holds 1 when the zero-cross data 5 output from the zero-cross detector 4 is 1, 0 when -1 and 0 when 0. Output. The digital phase comparator 8 receives the audio square wave signal 7 and the square wave signal 14 output from the digital VCO (voltage control oscillator) 13 and compares the phases as shown in FIGS. As shown in FIG. 2, when the phase of the square wave signal 14 output from the digital VCO unit 13 is behind the phase of the audio square wave signal 7, the digital phase comparison unit 8 outputs a phase delay signal 9. . As shown in FIG. 3, when the phase of the square wave signal 14 output from the digital VCO unit 13 is advanced compared to the audio square wave signal 7, the digital phase comparison unit 8 outputs the phase advance signal 10. . The integrator 11 integrates the phase lag signal 9 and the phase advance signal 10 output from the digital phase comparator 8 according to the following equation 2, and outputs a digital voltage value 12.

[0010]

## EQU2 ## where A _l = A _{l -1} + B _{l -1} + C _{l -1} where the voice pitch of 100 Hz is the sampling frequency.
8 kHz, 1 point sampling (1 is 1 to 80)
I do. Where A _l is the digital voltage value at the l-th point and 12 B _l−1 is the phase voltage at the l−1-th point.
The corresponding digital phase comparator voltage value C _l-1 corresponds to the phase advance at the 1-1 point.
The voltage value of the corresponding digital phase comparator .

The digital VCO unit 13 receives the digital voltage value 12, performs a voltage-frequency conversion having the characteristics shown in FIG. 4, and outputs a square wave signal 14.

In the voice pitch extracting device configured as described above, since the digital phase comparing unit 8, the integrating unit 11, and the digital VCO unit 13 form a PLL (phase locked loop), the digital voltage value 12 is It changes following the audio square wave signal 7. The frequency of the audio square wave signal 7 at that time is the digital voltage value 1 shown in FIG.
2 and can be calculated back from the relationship between the frequency of the square wave signal 14, A
_From the digital voltage value 12 corresponding to the value of ₈₀ , the voice pitch frequency is determined.

In the present embodiment, the product-sum calculation is only addition and subtraction of the equation (2) and the digital phase comparison calculation as compared with the pitch extracting apparatus using a conventional correlator. Can be extracted. For example, when comparison is made under the same conditions as in the description of the prior art, that is, when extracting a voice pitch of 100 Hz under an environment of a sampling frequency of 8 Hz, the product-sum operation is required by the following equation (2).
Only when calculating By the way, as shown in FIG. 5, the number of data required for the PLL to lock is at least one cycle, and is therefore 80 points. Therefore, the required number of product-sum operations is 80 × 1 = 80, and the voice pitch period can be obtained with such a small number of calculations. Therefore, the number of product-sum operations is reduced to 80/12600 = 1/150 compared to the conventional pitch extractor using a correlator, and the voice pitch extraction accuracy is equivalent to that of a pitch extractor using a correlator. Is obtained.

[0014]

According to the present invention, a voice pitch can be extracted with an extremely small number of product-sum operations and with an extraction accuracy equal to or higher than the conventional one.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a voice pitch extraction device according to an embodiment of the present invention.

FIG. 2 is a waveform chart showing an operation of the digital phase comparator 8;

FIG. 3 is a waveform chart showing an operation of the digital phase comparator 8;

FIG. 4 is a diagram illustrating an operation of a digital VCO unit 13;

FIG. 5 is a waveform diagram showing a period required for a PLL to lock.

FIG. 6 is a block diagram showing a conventional voice pitch extraction device.

FIG. 7 is a diagram illustrating an example of an audio signal.

FIG. 8 is a diagram illustrating an example of first-order autocorrelation.

FIG. 9 is a diagram illustrating an example of an 80th-order autocorrelation.

[Explanation of symbols]

1; audio signal, 2; A / D converter, 3; digital data, 4; zero-cross detector, 5; zero-cross detection data, 6; square wave generator, 7; audio square wave signal, 8; digital phase comparator 9; phase lag signal; 10; phase advance signal; 11; integrating section; 12; digital voltage value; 13; digital VCO section; 14; square wave signal;

──────────────────────────────────────────────────続 き Continued on the front page (58) Fields surveyed (Int. Cl. ⁷ , DB name) G10L 11/00-21/06 JICST file (JOIS)

Claims (2)

(57) [Claims] An A / D converter for converting an audio signal into a digital audio signal, a zero-cross detector for detecting a change in the sign of the digital audio signal, and a square for generating an audio square wave signal from the change in the sign comparing the wave generator, and a digital VCO unit which receives the digital voltage and outputs a square wave signal having a frequency corresponding to the voltage value, the square wave signal and a phase of the audio square wave signal, the phase delay signal And a digital phase comparator for outputting a phase advance signal, and a value obtained by integrating the phase delay signal and the phase advance signal with the digital V
Measuring the output of the integrating section to be output to the CO section ;
Speech pitch extraction device, characterized by chromatic and terminals that. 2. The digital phase comparator, the integrator and the digital VCO form a phase-locked loop, whereby a digital voltage value following the frequency of the audio square wave signal is obtained. 2. The voice pitch extraction device according to 1.