Speech enhancement algorithm based on independent component analysis in noisy environment

Voice signal enhancement has broad application prospects, but the existing methods have limited effect in conference rooms or teleconferences in a strong noise environment. In order to enhance the speech effect, this paper proposes a speech enhancement algorithm based on time-delay estimation of mic...

Full description

Saved in:

Bibliographic Details
Published in:	2020 IEEE International Conference on Artificial Intelligence and Information Systems (ICAIIS) pp. 456 - 461
Main Authors:	Hao, Xueyuan, Shi, Yu, Yan, Xiaohong
Format:	Conference Proceeding
Language:	English
Published:	IEEE 01.03.2020
Subjects:	Entropy Feature extraction ICA feature extraction Independent component analysis microphone array Microphone arrays Speech enhancement Wiener filtering Wiener filters
Online Access:	Get full text
Tags:	Add Tag No Tags, Be the first to tag this record!

Description
Summary:	Voice signal enhancement has broad application prospects, but the existing methods have limited effect in conference rooms or teleconferences in a strong noise environment. In order to enhance the speech effect, this paper proposes a speech enhancement algorithm based on time-delay estimation of microphone sound source localization. The algorithm combines Independent Component Analysis (ICA) and Wiener filtering. The algorithm uses the negative entropy of fast independent component analysis algorithm (FastICA) to extract feature and separate the speech signal, and then uses Wiener filtering to minimize the mean square error between the estimated signal and the speech signal extracted by the feature in the ICA domain. The paper deduces the unmixing matrix of the ICA transform in detail, and simulates the speech enhancement capability of the algorithm through Matlab. Simulation results show that the algorithm has obvious enhancement effect, and it can effectively reduce noise.
DOI:	10.1109/ICAIIS49377.2020.9194905