JPS5912199B2 - Audio parameter modification method - Google Patents

Description

Detailed Description of the Invention (1) Technical Field of the Invention The present invention provides a speech parameter correction method for automatically suppressing the occurrence of sounds that are difficult to hear, such as abnormal noises, in synthesized speech in a speech analysis and synthesis method. Regarding.

(2) Background of the technology Generally, in linear prediction methods such as PARCOR and LSP, speech is calculated using an all-pole model expressed by H(z)=1+ΣαiZ 1=1, where P is the prediction order and αi is the linear prediction coefficient. The spectral envelope characteristics of are approximate, so this model is often not applicable to nasal consonants such as N and M 10, plosive sounds, or vowel sounds, which have a low first formant frequency. When the fundamental frequency almost matches the first formant frequency, the bandwidth of the first formant may be analyzed to be abnormally narrow.

In such cases 15, the synthesized speech often has a very large amplitude or contains abnormal sounds that are difficult to hear. (3) Conventional technology and problems Conventionally, when analyzing and synthesizing speech, it is necessary to manually perform parameter abnormalities while repeating synthesis and listening in order to improve the quality of synthesized speech. I had to repeat the process of searching for a certain point in time, and then making appropriate modifications to the parameters.

Moreover, in the past, it was not always possible to accurately determine which part of the speech parameters caused the abnormality in the synthesized speech 95. If an abnormal sound such as "Giyon" is mixed into the synthesized speech, modifying the parameters is a matter of trial and error30
This resulted in inefficient processing. (4) Purpose of the Invention The present invention automatically detects and corrects the above-mentioned parameters that are particularly likely to cause abnormal sounds.
The purpose 35 is to improve the quality of synthesized speech.

(5) Structure of the Invention The present invention automatically detects abnormal parameters and corrects them, so when there is a high degree of proximity between a polar frequency with a narrow bandwidth in the spectrum envelope and a pitch frequency, the synthesized voice This was done by focusing on the fact that the power level in the vicinity of the polar frequency increases abnormally, causing abnormal noise.

FIG. 1 is an explanatory diagram of the relationship between the above-mentioned pole frequency and pitch frequency.

In the figure, F is a plurality of polar frequencies Fi and a bandwidth Bi
This is the spectral envelope approximately expressed by The present invention provides a polar frequency Fi (for example, 300 Hz) having a particularly narrow bandwidth Bi among the bandwidth Bi of each pole frequency.
If the pitch frequencies Pi are close to each other with a difference of, for example, several Hz to 30 Hz, it is determined that the parameter is abnormal,
This is to shift the pitch frequency by an appropriate value. The present invention is based on the above-mentioned principle and includes a speech analysis and synthesis method in which speech waves are analyzed, information on pitch frequency and information on spectral envelope are extracted, and speech synthesis is performed using this as a parameter time series. , means for approximating the information regarding the spectral envelope by a function of a polar frequency and a bandwidth; a means for extracting a polar frequency with a narrow bandwidth among the polar frequencies of the function; means for determining the proximity of the pitch frequency, and means for changing the pitch frequency when the existence of a polar frequency having a narrow bandwidth and close to the pitch frequency is detected by the extraction means and the determining means. Therefore, by widening the interval between the pole frequency and the pitch frequency where the bandwidth is narrow,
It is characterized by suppressing the occurrence of abnormal sounds in synthesized speech.

(6) Examples of the Invention The present invention will be described in detail below using examples.

FIG. 2 is a block diagram of a speech analyzer embodying the present invention.
In the figure, 1 is a speech analysis processing section, and 2 is a speech parameter correction processing section according to the present invention. In the speech analysis processing section 1, 3 is a sound source information analysis section that extracts pitch, amplitude, voiced/unvoiced information from input speech, and 4 is a linear predictive analysis and PAR for parameterizing spectral envelope information.
This is a linear predictive analysis unit that performs COR conversion. Analysis department 3,
The analysis output from 4 is used for speech synthesis in a speech synthesizer, not shown. In the parameter correction processing unit 2, 5 is the PARCOR output from the linear prediction analysis unit 4.
This is a conversion unit that converts coefficients back into linear prediction coefficients.

If linear prediction coefficients can be extracted from the linear prediction analysis unit 4 as intermediate processing data, the conversion unit 5 is not necessary. 6 is the transfer function H@) of the all-pole approximation model regarding the spectral envelope of the voice mentioned above, and the equation when the denominator is A(z) is expressed as A(z)= This is a root-finding calculation unit that finds roots on the z-plane that satisfy O.

Reference numeral 7 denotes a conversion unit that calculates the polar frequency and bandwidth, where the solution of the calculation result of the root-finding performance unit 6 is taken as the sampling period and T is the sampling period.

8 is a part for extracting the polar frequency Fi having a narrow bandwidth Bi.

The degree of narrowness of the bandwidth is determined and extracted depending on whether the value of Bi is less than a certain threshold value or more than a certain running force threshold value.

Reference numeral 9 denotes a determination unit that determines the degree of proximity between the extracted polar frequency of the narrow bandwidth and the pitch information (frequency) Pi being output from the sound source information analysis unit 3.

For example, the proximity O determination is performed using the following equation Fi 〉100 1fi-Pil It is carried out based on.

When a polar frequency Fi that satisfies this determination formula is found, it is determined to be an abnormal parameter and output. Reference numeral 10 denotes a pitch frequency changing unit, which corrects the pitch frequency so that the pitch frequency is separated by a certain value or more, for example, 30 Hz or more, with respect to the polar frequency having a narrow bandwidth, and supplies it to the analysis output as Pi. .

In addition to modifying the pitch frequency, it is also possible to perform modifications such as widening the bandwidth of the polar frequency and adjusting the time window.

(7) Effects of the Invention The present invention can automatically correct in advance parameters that are likely to cause abnormal speech, making it extremely easy to improve the quality of synthesized speech. .

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is an explanatory diagram of the principle of the present invention, and FIG. 2 is a configuration diagram of an embodiment of the present invention. In the figure, 1 is a speech analysis processing section, 2 is a speech parameter correction processing section, 3 is a sound source information analysis section, 4 is a linear prediction analysis section, 5 is a conversion section from PARCOR coefficients to linear prediction coefficients,
6 is a root calculation unit, 7 is a conversion unit from roots to pole frequency and bandwidth, 8 is a narrow bandwidth extraction unit, 9 is a proximity determination unit between pole frequency and pitch frequency, and 10 is a pitch frequency change unit. , respectively.

Claims (1)

[Claims] 1 In a speech analysis and synthesis method that analyzes speech waves, extracts information regarding pitch frequency and information regarding spectral envelope, and performs speech synthesis using this as a parameter time series, the information regarding the spectral envelope is combined with polar frequencies and bandwidth. means for approximating by a function; means for extracting one with a narrow bandwidth among the polar frequencies of the function; means for determining the proximity between the polar frequencies of the function and the pitch frequency; means for changing the pitch frequency when the existence of a pole frequency having a narrow bandwidth and close to the pitch frequency is detected by the determining means, the pole frequency having the narrow bandwidth and the pitch frequency; A speech parameter modification method characterized by suppressing the occurrence of abnormal sounds in synthesized speech by widening the interval between.