Method for carrying out blind source separation on convolutionary aliasing voice signals

Abstract

The invention provides a method for carrying out blind source separation on convolutionary aliasing voice signals. Firstly, a time domain convolutionary aliasing model is converted into a frequency domain multi-channel linear instantaneous convolutionary aliasing model, which can be realized by the following steps: firstly, converting convolutionary aliasing time domain signals into a frequency domain; then carrying out relatively independent ICA operations on each channel to obtain independent components. Next, the independent components are rearranged by an MSBR algorithm, which specificallycomprising the following steps: firstly, classifying signals of different frequency bands; then progressively obtaining transposed matrixes according to different object functions step by step, wherein the steps of rearrangement are mutually complementary. The MSBR algorithm utilizes the strong relevance of harmonic frequency to improve the iteration accuracy and solves the residual uncertainty of residual frequency bands according to the continuity of adjacent frequency bands and corresponding reference frequencies, and the computational complexity of the MSBR algorithm is approximately in direct proportion to the number of reference frequency bands. The invention improves the convergence efficiency and the accuracy, is more suitable for real-time processing, has good separation performance of convolutionary mixed voice signals and can also be applied to real phonetic environment.

Description

technical field [0001] The invention relates to a method for blind source separation of convolutional and aliased speech signals in a multiple-input multiple-output (MIMO) system without channel state parameters, which can be widely used in neural networks and multi-antenna systems, especially in speech signal processing aspect. Background technique [0002] Blind source separation (BSS) of speech signals is a recent research hotspot. The real speech environment can be approximated as a convolutional aliasing model, so higher requirements are put forward for the convolutional aliasing speech signal BSS. [0003] Traditional convolution and aliasing speech signal BSS algorithms can generally be divided into two categories: [0004] 1. Deconvolution is performed directly in the time domain; [0005] 2. Transform to other transform domains such as wavelet domain or frequency domain for processing. [0006] Since there may be many filtering coefficients, the first type of alg...

AI Technical Summary

Problems solved by technology

[0014] Aiming at the uncertainty problem of the existing voice signal BSS, the present invention provides a method for blind source separation of convolutional and aliased voice signals that can improve the separation performance of the MIMO system under the condition of no channel state information

Images