A time efficient architecture implementation of PCA for ICA

Abstract

Independent component analysis (ICA) Algorithm is a popular algorithm for BSS (Blind Source Separation) and Principal component analysis (PCA) works as its preprocessing algorithm. This work proposes time efficient and high precision PCA architecture which operates on 2 channels each with 3000 samples in single precision FP (floating point) format. Architecture uses high speed parallel computations for throughput producing more than 3.5 Gflops and FP units are incorporated for high precision.