Voice separation method and device, electronic equipment and storage medium

Abstract

The invention relates to the technical field of computers, in particular to the technical field of voice processing, and provides a voice separation method and device, electronic equipment and a storage medium, which are used for improving the accuracy of voice separation. The method comprises the following steps: acquiring to-be-separated voice data; encoding the to-be-separated voice data to obtain a corresponding encoding result, and performing voice separation operation on the encoding result to obtain voice features and phase movement features corresponding to the at least one target sound source; based on the obtained at least one phase movement feature, performing parameter reconstruction on a preset signal decoding parameter set to obtain a corresponding target signal decoding parameter set; and decoding the corresponding voice features by using the obtained at least one target signal decoding parameter set to obtain reconstructed voice signals corresponding to the corresponding target sound sources. Since the modeling capability of the medium-high frequency signals is improved through phase shifting, almost no additional parameter quantity or complexity is added, and the separation performance can be effectively improved.

Description

technical field [0001] The present application relates to the field of computer technology, in particular to the technical field of speech processing, and provides a speech separation method, apparatus, electronic device and storage medium. Background technique [0002] Related end-to-end speech separation systems widely use an encoder-separator-decoder structure, where the encoder part is used to transform the input audio signal (mixture waveform signal) to obtain richer features, and the separator The features generated by the encoder are received for a separation operation, and the decoder reconstructs the output of the separator into a waveform signal. [0003] In the related art, the encoder and decoder parts, in most end-to-end separate systems, are two linear transform parts. When optimizing the network parameters of the encoder and decoder, the method of random initialization + joint optimization is often used, that is, random initialization before the network train...

AI Technical Summary

Problems solved by technology

[0004] However, the jointly optimized encoder and decoding parameters have the characteristics of focusing too much on low-frequency signals, and there is a phenomenon of repeated modeling of different phase features of the same magnitude in the low-frequency part, which reduces the performance of mid- and high-frequency signals. modeling capabilities, and the parameter utilization of encoders and decoders

Images