Thesis Archive
An Implementation Of A PC-Based Blind Source Separation Using the Information Maximation Approach (2007)
Raymund Patrick Z. Ramos, Ann Jacqueline S. Tan and Ma. Karina B. Villanueva, advised by Engr. Edzel R. Lapira
Abstract:
-Blind Source Separation has caught the attention of many in the field of digital signal processing. Various algorithms have already been presented as solution to the much celebrated Cocktail Party Problem. One of those, called Information Maximization approach proposed by Anthony J. Bell and Terrence J. Sejnowski, will be the focus of this research. The aforementioned algorithm is implemented in MATLAB 6.5. Two speech signals, captured by using two soundcards, are combined with an unknown linear mixing process. The INFOMAX algorithm, having no prior knowledge about the input signals, maximizes the mutual information through gradient adaptation in order to generate the appropriate demixing matrix such that the two statistically independent signals would be recovered.
Correspondence:
Engr. Edzel R. Lapira
[email protected]