An FPGA-based implementation of fastica for variable-length 4-channel signal separation

Abstract

Independent Component Analysis (ICA) is one of the most popular and powerful tool that has been used widely in the field of signal processing. Due to its complexity, implementing ICA became a challenge for designers. In this paper, authors proposed an FPGA-based ICA implementation using FastICA algorithm. The design can process 4 audio channels with variable length from $2^9$ to $2^{25}$ samples. The proposed implementation achieves the speed of 11.27 Mbps and can process over 1.4 million samples per second.