<abstract xml:lang="en" xmlns="http://www.w3.org/1999/xhtml"><p>The main focus of the paper is to bring out the differences in performance related issues of Fast-ICA algorithm associated with floating point and fixed point digital signal processing (DSP) platforms. The DSP platforms consisting of TMS320C6713 floating point processor and TMS320C6416 fixed point processor from Texas Instruments have been chosen for this purpose. To study the consistency of performance, the algorithm has been subjected to three different test cases comprising of a mixture of synthetic signals, a mixture of speech signals and a mixture of synthetic signals in presence of noise, respectively. The performance of the Fast-ICA algorithm on floating point and fixed point platform are compared on the basis of accuracy of separation and execution time. Experimental results show insignificant differences in the accuracy of separation and execution time obtained from fixed point processor when compared with those obtained from floating point processor. This clearly strengthens the feasibility issue concerning hardware realization of Fast-ICA on fixed point platform for specific applications.</p></abstract>

The main focus of the paper is to bring out the differences in performance related issues of Fast-ICA algorithm associated with floating point and fixed point digital signal processing (DSP) platforms. The DSP platforms consisting of TMS320C6713 floating point processor and TMS320C6416 fixed point processor from Texas Instruments have been chosen for this purpose. To study the consistency of performance, the algorithm has been subjected to three different test cases comprising of a mixture of synthetic signals, a mixture of speech signals and a mixture of synthetic signals in presence of noise, respectively. The performance of the Fast-ICA algorithm on floating point and fixed point platform are compared on the basis of accuracy of separation and execution time. Experimental results show insignificant differences in the accuracy of separation and execution time obtained from fixed point processor when compared with those obtained from floating point processor. This clearly strengthens the feasibility issue concerning hardware realization of Fast-ICA on fixed point platform for specific applications.

Implementation of Fast-ICA: A Performance Based Comparison Between Floating Point and Fixed Point DSP Platform

Electrical Engineering, Indian Institute of Technology, 721302, Kharagpur, India

I. T. E.R, Electronics and Communication Engineering, SOA University, 751030, Bhubaneswar, India

Measurement Science Review

The main focus of the paper is to bring out the differences in performance related issues of Fast-ICA algorithm associated with floating point and fixed point...

Implementation of Fast-ICA: A Performance Based Comparison Between Floating Point and Fixed Point DSP Platform

Publié en ligne: 21 sept. 2011

Pages: 118 - 124

DOI: https://doi.org/10.2478/v10048-011-0022-1

Mots clés
blind source separation, fast-ICA, floating point implementation, fixed point implementation

This content is open access.

Implementation of Fast-ICA: A Performance Based Comparison Between Floating Point and Fixed Point DSP Platform

Publié en ligne: 21 sept. 2011

Pages: 118 - 124

DOI: https://doi.org/10.2478/v10048-011-0022-1

Mots clésblind source separation, fast-ICA, floating point implementation, fixed point implementation

This content is open access.

Mots clés
blind source separation, fast-ICA, floating point implementation, fixed point implementation