Publication

Abstract : In this article, we study Fast Independent Component Analysis (FastICA) algorithm for resolving closely spaced sinusoids embedded in noise. We generated closely spaced sinusoid signals embedded in noise using open source IT++ libraries. The closely spaced sinusoids signals were randomly generated with minimum frequency separation of [0.05 * F s : 0.2 * F s ]. Simulations were performed to resolve closely spaced sinusoids using FFT algorithm with minimum frequency of seperation [0.05 * F s : 0.2 * F s ], but FFT failed to resolve closely spaced sinusoid of such lower frequencies and produced more error for multiple signals and different resolutions. After a through analysis, we studied FastICA to resolve the problem, and has given good results. The performance of FastICA algorithm is evaluated for different sampling rates as well as different number of signal components as done with FFT. It is concluded that FastICA has better performance compared to Fast Fourier Transform (FFT) for resolving multiple closely spaced signals in the presence of Additive White Gaussian Noise (AWGN). The entire simulation results were generated using IT++ libraries.