Title :
The Research of Blind Signal Separation Algorithm Based on Neural Network
Author :
Liu, Hongjie ; Feng, BoQin ; Zheng, Hongming
Author_Institution :
XVan Jiaotong Univ., Xi´´an
Abstract :
In order to solve the problem of bad separation result of Anti-Hebbian algorithm, a weighted algorithm of blind signal based on the feed back neural network is discussed, i.e. the influence of source signal is overcame by the weighted process. And furthermore, the similarity algorithm is presented. The main process of the blind signal separation algorithm is to construct neural network by the characteristic of blind signal, then to train network using the weighted algorithm or similarity algorithm and get the result at last. The results of the simulation experiment and the practical speech signal separation show the separation effect of the weighted algorithm and similarity algorithm is superior to that of Anti-Hebbian algorithm.
Keywords :
blind source separation; learning (artificial intelligence); recurrent neural nets; speech processing; antiHebbian weighted algorithm; blind signal separation algorithm; feedback neural network training; similarity algorithm; speech signal separation; Blind source separation; Feeds; Gaussian distribution; Neural networks; Signal detection; Signal processing; Signal processing algorithms; Source separation; Speech; Statistics;
Conference_Titel :
Natural Computation, 2007. ICNC 2007. Third International Conference on
Conference_Location :
Haikou
Print_ISBN :
978-0-7695-2875-5
DOI :
10.1109/ICNC.2007.738
