A CCA Criterion Based Adaptive Algorithm for Blind Extraction of Specific Signal

Authors

  • Wei-Tao Zhang, Xiao-Guang Yuan, and Shun-Tian Lou Author

Keywords:

Blind source separation, canonical correlation analysis, steepest descent, Newton iteration.

Abstract

In this paper, the canonical correlation approach 
for blind source separation is revisited. We proposed a novel 
criterion for source extraction, which is proved to be equivalent 
to the CCA criterion, whereas it has some advantages over the 
standard CCA criterion in developing fast learning algorithms. 
By optimizing the proposed criterion, we developed two 
learning algorithms for the extraction of specific signals. The 
first algorithm is based on the steepest descent technique, and 
the second one is a modified Newton algorithm. The experiment 
results demonstrate the effectiveness of the proposed learning 
algorithms, and show that the modified Newton algorithm 
converges much faster than the other extraction method. 

Downloads

Download data is not yet available.

Downloads

Published

18.09.2014

How to Cite

A CCA Criterion Based Adaptive Algorithm for Blind Extraction of Specific Signal. (2014). International Journal of Information and Electronics Engineering, 4(5), 370-374. https://ijiee.org/index.php/ijiee/article/view/541