FPGA implementation of the ICA algorithm using multiplexing

Abstract

This work presents an optimized version in FPGA technology of a digital system, which solves in real time the Blind Source Separation problem using the Independent Component Analysis, ICA algorithm and following the Maximum Information technique, INFOMAX. To demonstrate the FPGA realization, we use a mix of three sinusoidal signals, which represents three independent sources, with 1000Hz, 800Hz and 600Hz values in frequency. The mixed signal is treated by the ICA system. The digital system in FPGA was analyzed first in Simulink of Matlab, evaluating its performance. Then, the FPGA architecture, which was optimized observing a multiplexing scheme, is proposed where the number of used DSP resources is minimal. This leads to extend this multiplexing scheme to cover future designs with more signals.