Human voice separation method and device, user terminal and storage medium

Abstract

The invention discloses a human voice separation method and device, a user terminal and a storage medium, and relates to the technical field of audio processing. The human voice separation method comprises the steps: a sampled to-be-separated audio file sound channel is separated to obtain an initial waveform sequence; the initial waveform sequence is subjected to discrete Fourier transform to obtain an initial two-dimensional array; the initial two-dimensional array is subjected to module obtaining to obtain an initial speech spectrum; the initial two-dimensional array is subjected to phase obtaining to obtain an initial phase image; the initial speech spectrum is guided into a convolutional neural network model to obtain a mask through calculation; the mask and the initial phase image are subjected to first point multiplication operation to obtain a human voice source speech spectrum; the human voice source speech spectrum and the initial phase image are subjected to second point multiplication operation; the result of the second point multiplication operation is subjected to inverse discrete Fourier transform to obtain single human voice source audio waveforms; and the single human voice source audio waveforms are spliced to obtain stereo audio. According to the human voice separation method and device, the user terminal and the storage medium, automatic human voice separation of the audio can be achieved.

Description

technical field [0001] The present invention relates to the technical field of audio processing, in particular to a human voice separation method, device, user terminal and storage medium. Background technique [0002] Usually for popular music, the human voice is the main theme, and the accompaniment is the rhythm of the music. Since the human voice is usually accompanied by background music, vocal separation is a challenging task. A prerequisite for musical instrument classification, and these techniques can be used in applications such as recommendation systems and label classification. One of the commercial applications of the vocal separation system is karaoke, meaning a musical track without the vocals. Karaoke music helps music lovers learn to sing an existing piece or sing it in a concert. Currently, extracting karaoke music is done during the recording process, which requires a lot of manual operations and time. [0003] Most of the existing deep learning technol...

AI Technical Summary

Problems solved by technology

Currently, extracting karaoke music is done during the recording process, which requires a lot of manual work and time

[0003] Most of the existing deep learning technologies for human voice separation improve the separation effect at the cost of reducing the sampling rate and reducing the number of channels. After separation, the audio quality is degraded and the auditory effect of the audio is reduced.

Images