Application of Blind Signal Processor for Signal Separation

• Main Authors: Yao-Yi Wang, 王耀毅
• Other Authors: Chuen-Yau Chen
• Format: Others
• Language: zh-TW
• Published: 2007
• Online Access: http://ndltd.ncl.edu.tw/handle/32132098826070492971

碩士 === 國立雲林科技大學 === 電機工程系碩士班 === 95 === In blind signal separation, the goal is to recover the source signals from the mixing signals which come from unknown environment. In recent years, adaptive blind signal separation utilizes some important techniques such as pre-whitening, post-whitening, natural gradient, orthogonality and de-correlation, which make blind signal separation algorithm working more efficiency. Then the output signals would very similar to the source signals. In this thesis, there are two methods to improve blind signal separation algorithm, utilizing the projection between output signals and input signals to correct separated matrix. These methods take less iteration and speed up the system to recover the souorce signals. In sub-gaussian experiment, these methods take 800~2300 iterations. In super-gaussian experiment, these methods improve denoise ability.