JPS584200A - Musical interval correction system - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention uses a method for correcting the pitch when synthesizing fp4 data and outputting a sound of 6-kJy. information (e.g. voice quality,
This article is about the Tsuji 1 Chōuramasa method, which preserves the pitch (with a slight difference of 1) and corrects only the pitch before outputting it.

As an i-voice synthesis method, nine PARCOR (Partlal autoco
rrrelation+ (partial self-phase 5lI) method, LPC
(linear pr*dietivecoding)
(Linear predictive coding) method, LSP (tinsspe
ctrum palr) (line spectrum pair) method, etc. are known. Speech synthesis is performed by, for example, converting each frame period manually, adding the interpolated output for each sampling period using linear interpolation etc. as a coefficient of a digital filter, and adding the pulse train or white noise from the sound source to the digital filter to perform speech synthesis. However, in the past, it was common for the uttered round sound and its background background sound, accompaniment sound, etc.) to be input at the same time, and the uttered sound and the background sound were input at the same time. Although each sound may be digitized when the sounds are extracted from separate sound sources and synthesized, it is customary to mix them as analog sounds when synthesizing them.

Therefore, considering the differences in pitch perception among individuals, it is inevitable that pitch deviations will occur even if the voice is uttered in time with the accompaniment, for example.

The present invention eliminates such misalignment, and 11! The purpose of this invention is to provide a sound type correction method that synthesizes information from different sound sources without impairing the quality of people's voices.

In order to achieve the above object, the present invention provides analysis means for separating input speech into vocal tract resonance characteristics and excitation sound source i′# information, and means for inputting pitch information different from the pitch of the first human-powered speech. a means for correcting pitch information in the excitation sound source information analyzed from the p@ki human voice using the f interval information; and an output for synthesizing and outputting a voice based on the excitation sound source information and the resonance characteristics of the vocal tract. The present invention is characterized in that it corrects only the f-root of human input speech and outputs it.

The present invention will be explained in detail below using several examples. FIG. 1 shows the configuration of the IT correction method according to the present invention, and in this embodiment, the PARCOR method is used. In the figure, M is microphone, AMPI.

AMP is an amplifier and filter circuit, AMPs is an amplifier, S is a speaker, IN is an accompaniment device (such as a stereo tape recorder), AN is a voice analysis section, CON is a voice synthesis section, F is a pitch frequency extractor, P is a pitch information register, and MIX is a mixer.

The voice analysis section AN includes a digital filter circuit as a PARCOR analysis section, an amplitude/voiced/unvoiced information extraction section T1 code 1 hee! It consists of 5 parts, an audio synthesis part CONFi decoder, an impulse generation circuit, a digital filter, etc.
In l~, is a parameter coefficient, ■ is voiced/unvoiced information 4, A is amplitude information, and N is pitch information.

A human-powered accompaniment device IN, for example a tape recorder, stores pitches uttered by a professional singer, etc. (pitch frequency 1 conditions created in advance in synchronization with tf-n) along with the accompaniment. The voice of the speaker is inputted through the microphone M into the sound 1iif + nail 81 ANK through the microphone M, and is passed through a digital filter to the para 4-p prefectural news and the wave width. , voiced and unvoiced information), and the amplitude, voiced and unvoiced information extractor T extracts pitch information of the speaker, line drop amplitude information, and voiced and unvoiced information Vt.
Take it out. The pitch information is frequency information CF8K)2t of pitch information synchronized with the accompaniment from the tape recorder IN.
Pitch information register PK via V-te waste wave number extractor Ft
The voice synthesizer CON uses the amplitude, voiced/unvoiced information, etc. of the speaker. Then, by using only the pitch information N inputted from the tape recorder, the synthesized sound 3 with the voice corrected is obtained.
By adjusting the level of the accompaniment tone 1 with t and the output of the beaker with the electric accompaniment tone 1 and the electric Φ with Mlx, the pitch can be corrected without changing the sound quality of the speaker.

As explained above, the method of the present invention, for example, when singing along with accompaniment, can preserve the timbre of the vocalist while correcting the tone of the pitch, and can use the sound 1i (pitch frequency) information for this purpose. may be created in advance in synchronization with the accompaniment and stored on the same tape as the nine-bot accompaniment. For example, even if you utter a completely different pitch, you can sing along with the accompaniment at the same pitch as if it were uttered by a professional singer.

The sound cage information is approximately 10m to 3Qm depending on the accompaniment sung by a professional singer etc.
Each @@@ is obtained as 7 bttii digital information. For example, use this as F8K (7r@gu@n@F8hi
ft K@F1nF) etc.) can be used to analyze the voice band.

If the analysis and synthesis method is a so-called vocoder method that separates sound source information and vocal tract transfer characteristics, then Percoll,
11 systems such as i, 8Pt, 7 ormanto are possible. Regarding the analysis section, compared to the normal pocoder system, rounding is possible without the need for a pitch extraction section, and is possible with a much simpler structure.

4. F! jAijo easy aIjI! The drawing is a block diagram of an embodiment of the pitch correction method of the present invention.

M: microphone, AN: voice analysis unit, T: amplitude/voiced/unvoiced information extractor, F: pitch information extractor, P: pitch information register, INi accompaniment input device, Mix.

z middle, S+speaker, K summer ~ D; voice parameter coefficient, M: amplitude information, v: voiced and unvoiced information, N: pitch information.

Claims (1)

[Claims] analysis means for separating the input voice into resonance characteristics and excitation i# source T information; means for manually generating pitch information different from the pitch of the input voice; f7! by means of correcting the f type information in the excitation sound source information analyzed by WjJ and the resonance characteristics of the vocal tract. ! and output means for synthesizing and outputting the pitch, and the pitch correction method has an n value of correcting and outputting only the pitch of the input voice.