AWGN信道中盲源分离的时频联合两步消噪预处理方案.pdf

30 4 V o l30 4 2009 4 A pr2009 Journal of H arb in Engineering U niversity AWGN 王尔馥, 张乃通, 孟维晓 哈尔滨工业大学 通信所, 黑龙江 哈尔滨 150080) : AWGN , 2 , , -, . , , , . : ; ; ; : TN 911. 7 : A : 1006-704 3 2009) 04-0390-05 A concurrent time-frequency andnoise-pretreatmentmethodology forblind source separation in an additivewhiteGaussian noise channel W ANG Er-fu, ZHANG Na-i tong, M ENG W ei-x iao Communication R esearch Center, H arb in Institute of T echnology, H arb in 1500 80, Ch ina) Abstract:The b lind source separation BSS) problem in additive wh ite Gaussian no ise AWGN ) is known as a hard problem; performance of the separation algorithm degrades significantly w ith decreases in the SNR. Emp irical mode de- composition EMD) and wavelet transform help so lve the problem. To comb ine the advantages of these two m ethods, the idea of rough cance llation and then exact cance llationw as adopted and a concurrent tmi e-frequency and no ise-pre- treatm ent m ethodology for BSS w as proposed. Smi ulation results show ed that this scheme decreased the additive no ise in observed signa ls, m aking the present blind source separation algorithm work w ith a larger range of SNR s and enhanc ing its robustness. M eanwhile, the necessity of data ex tension to the communicat ion and signalw as shown. Keywords: blind source separation; no ise-pretreatm ent; w avelet transform; em pir ical m ode decomposition b lind source separation, BSS) . , , , . . , . Cardoso , FOBI) [ 1] , , .