Method and device for removing accompaniment from song

Abstract

The invention provides a method for removing an accompaniment from a song. The method comprises the steps that an accompaniment audio signal and a song audio signal are obtained; FFT is carried out on the song audio signal and the accompaniment audio signal to obtain a song audio signal amplitude spectrum, an accompaniment audio signal amplitude spectrum and the phase of the song audio signal; the accompaniment audio signal amplitude spectrum is enhanced; the enhanced accompaniment audio signal amplitude spectrum is subtracted from the song audio signal amplitude spectrum to obtain a result, inverse FFT is carried out with the combination of the phase and the result, and therefore an audio signal without the accompaniment is obtained. The invention further provides a device for removing the accompaniment from the song. According to the method and device for removing the accompaniment from the song, the degree of purity of the song obtained after separation is accomplished can be improved, separation execution efficiency is high, and an algorithm is simple and easy to implement.

Description

technical field [0001] The invention relates to the field of audio signal processing, in particular to a method and device for accompaniment of songs. Background technique [0002] Singing separation system has wide applications in some fields, such as automatic recognition and correction of lyrics. Automatic lyrics recognition usually requires that the input system is a separate singing voice, that is, only singing voice without accompaniment, but this is often not practical for almost all songs, because most songs contain both vocal and instrumental accompaniment. [0003] At present, there is very little research on the separation of singing voices in music. In the problem of separating voices from different sound sources, such a task is easy for humans, but difficult for machines. Speech separation has been extensively studied, but since music is an extremely complex signal, multiple signals including singing voices and different musical instruments are mixed together, ...

AI Technical Summary

Problems solved by technology

The separation process mainly depends on the characteristics of the main pitch. In many cases, there will be overlap between the fundamental tone and overtone of the singing voice and the instrument. It is difficult to fully obtain the TF information of the singing voice by using the main pitch alone, so the singing voice and the accompaniment are often still inseparable.

Moreover, this method has problems such as complex algorithm, large amount of calculation, and low execution efficiency.

Images