Analysis of source separation algorithms in industrial acoustic environments

Abstract

This paper shows the results from the computation cost evaluation of three blind source separation algorithms. The algorithms tested were: FastICA, Adaptive Algorithm Based on Natural Gradient, and Adaptive EASI Based on Relative Gradient. The algorithms were chosen for their relative simplicity, and taking into account their hardware implementation feasibility, either on a FPGA or an ASIC, as part of a system for acoustic localization of mobile agents in industrial environments.