A Single-Channel Speech Blind Separation Method Based on Computational Auditory Scene Analysis

Abstract

The invention relates to a single-channel speech blind separation method based on computational auditory scene analysis. The method includes the following steps: step 1, performing front-end processing on the input aliasing speech; step 2, processing the aliasing speech processed in step 1 Perform auditory feature extraction; step 3, perform auditory segmentation on the aliasing speech processed in step 2 based on harmonic characteristics; step 4, perform auditory reorganization on the aliasing speech processed in step 3 based on energy features; Four processed aliased speech for target speech reconstruction. In the presence of noise, the present invention can well solve the problem of single-channel aliasing speech separation, and the separated speech can be applied to the front end of speech recognition, and will have wide applications in the fields of artificial intelligence, speech communication, and sound signal enhancement prospect.

Description

technical field [0001] The invention relates to a single-channel speech blind separation method based on computational auditory scene analysis, which belongs to the field of speech signal processing. Background technique [0002] Speech signal processing is one of the most eye-catching research fields in signal processing in recent years. In the field of speech signal processing, an important problem is how to separate the original speech from the aliased speech obtained by random mixing of unknown signals. The aliased speech Separation has important research significance and practical value in noise elimination, speech synthesis, speech communication and so on. Due to the overlap of aliased speech in the time domain and frequency domain, it is difficult for the commonly used speech enhancement methods to achieve speech separation. At present, the most commonly used method for aliasing speech separation is the blind source separation method, that is, the independent compone...

AI Technical Summary

Problems solved by technology

Due to the overlap of aliased speech in time domain and frequency domain, common speech enhancement methods are difficult to achieve speech separation

At present, the most commonly used method for aliasing speech separation is the blind source separation method, that is, the independent component analysis method, but the independent component analysis method has some assumptions, such as requiring that the number of observed signals is greater than or equal to the number of source signals, assuming no Noise, etc., these conditions limit the application of blind source separation, therefore, blind source separation cannot be applied to the single-channel aliasing speech separation problem where the number of source signals is greater than the number of observation signals, especially in the presence of noise, blind source separation The effect of the separation method applied to the separation of aliased speech is obviously deteriorated

Images