Concatenation point smoothing method in waveform concatenation, device and storage medium

Abstract

The invention relates to the field of voice signal processing, and provides a concatenation point smoothing method in waveform concatenation. The concatenation point smoothing method in the waveform concatenation is applied to an electronic device. The concatenation point smoothing method in waveform concatenation comprises the steps that concatenation points of two voice units to be subjected toconcatenation are determined, and voice signal segments with preset lengths at the two concatenation points are intercepted correspondingly; the two voice signal segments are correspondingly subjectedto windowing processing through a window function to correspondingly obtain corresponding short-time analysis signals; the amplitudes, phases and frequency of the two short-time analysis signals areacquired correspondingly based on the short-time Fourier transform; polynomial interpolation based on distance weight is carried out on the amplitudes, phases and frequencys of the two short-time analysis signals to acquire new amplitudes, phases and frequency; and sine wave synthesis is carried out on the new amplitudes, phases and frequency to acquire new voice signal segments. According to theconcatenation point smoothing method in waveform concatenation, data of the voice units are analyzed through a sinusoidal model, and the voice signals at the concatenation position are expressed as the sum of a series of sine waves, so that the smooth transition of synthesized voice is ensured, and the improvement of the naturalness degree of the synthesized voice is facilitated.

Description

technical field [0001] The present invention relates to the technical field of speech signal processing, in particular to a splicing point smoothing method, device and computer-readable storage medium in waveform splicing. Background technique [0002] Waveform splicing technology is a technology applied in the speech synthesis system. This technology synthesizes the required voice by splicing pre-recorded unit voice data. Among them, the PSOLA technology is a pitch-synchronous speech analysis / synthesis technology, which first requires accurate pitch period and determination of its starting point. The judgment error of the pitch period or its starting point will affect the effect of PSOLA technology. Secondly, PSOLA technology is a simple splicing synthesis of waveform mapping. Whether this splicing can maintain a smooth transition and its impact on frequency domain parameters has not been resolved. [0003] In addition, the unit speech data used for splicing often have d...

AI Technical Summary

Problems solved by technology

The judgment error of the pitch period or its starting point will affect the effect of PSOLA technology

Secondly, PSOLA technology is a simple splicing synthesis of waveform mapping. Whether this splicing can maintain a smooth transition and its impact on frequency domain parameters has not been resolved.

[0003] In addition, the unit speech data used for splicing often have differences in frequency or pitch. Therefore, after splicing using the TD-PSOLA algorithm, it will bring discontinuity in the synthesized speech spectrum, and the pitch correction is relatively large. Sometimes this discontinuity will be very obvious, manifested as a karaoke sound in the synthetic voice

Images