Phoneme changing method based on digital signal processing

Abstract

This invention published a kind of voice variation method that based on digital signal processing. It includes the steps of (1) selecting the original voice signal that needs to change; (2) finding the basic tone cycle length of original voice signal; (3) confirming the position of every basic tone cycle of whole original voice signal according to the basic tone cycle length; (4) inserting or deleting the basic tone cycle between the basic tone cycle of original voice signal, get the shortened or prolonged voice signal; (5) linearly extending or compressing the shortened or prolonged voice signal to the same length of original voice signal and getting the changed voice signal. This invention can be realized real-time on DSP chip. The changed voice is very natural.

Description

technical field [0001] The present invention relates to a voice changing method, more specifically, the present invention relates to a voice changing method based on digital signal processing. Background technique [0002] Fundamental frequency and formants are two very important features in speech. The fundamental frequency is the frequency at which the vocal cords vibrate when making voiced sounds. The fundamental frequency is directly related to the gender of the speaker. Generally speaking, the fundamental frequency of a male voice is relatively low, and that of a female voice is relatively high. In addition, age also has a certain influence on the base frequency. The base frequency of the elderly is lower than that of the young, and the base frequency of the young is lower than that of children. Therefore, by changing the fundamental frequency, the effect of speech can be changed, affecting people's judgment on the speaker's age and even gender. [0003] Formant refer...

AI Technical Summary

Problems solved by technology

The common problems of these methods are that the amount of calculation is relatively large, manual intervention is required, and the naturalness of the synthesized speech is relatively poor.

However, due to the limitations of the PSOLA algorithm itself, when the fundamental frequency needs to be changed in a relatively large range, the speech will generate quite a lot of aliasing, resulting in a lot of noise

After the other two methods change the fundamental frequency, the naturalness of the voice is worse, and the calculation amount of these two methods is relatively large, and it will be difficult to realize it in real time on the DSP chip.

In addition, these methods will change the length of the original speech, which will cause a big problem for the real-time transmission of the changed speech

And what existing voice changing method basically all adopted is the method for analog circuit, is not suitable for realizing on digital signal processor (DSP)

Images