"Studies on effectiveness of the JADE algorithm" by Theresa Lye and Prof. Athina Petropulu
Blind signal separation is the process of separating signals from one another by observing only their mixtures. This procedure can be applied to image and sound processing, communications, and to biomedical signals such as EEG signals.
There are various methods of accomplishing blind signal separation. Independent component analysis (ICA) is one method that exploits the non-Gaussianity of independent signals. The JADE algorithm is an implementation of ICA based on the joint diagonalization of matrices.
I tested the effectiveness of the JADE algorithm on various types of signals and developed a method intended to correct the permutation and amplitude ambiguity present in ICA. I also attempted to window non-stationary signals into stationary parts so that the signals may be properly separated by the JADE algorithm.
Theresa Lye is in the Class of 2013 and is pursuing a major in ECE and a minor in Computer Science. She has been working under Dr. Petropulu since Fall 2010