Blind Source Separation in the Presence of AWGN Using ICA-FFT Algorithms a Machine Learning Process
M. R. Ezilarasan, G. Rajesh, Kumar R. Vinoth, K. Aanandhasaravanan · 2023
The process of isolating sound sources in any audio environment is known as blind source separation (BSS) or audio signal separation. Applications for source signal separation include signal processing, audio processing, etc. This study investigates how to distinguish sound signals from a mixture when there is additive white Gaussian noise (AWGN). The independent component analysis (ICA) method is unable to precisely assess separated signals as a result of this additive noise. In this article, a mixing matrix is employed, which downsamples several sound sources before combining them. The mixed signal is given a variable amount of AWGN as input. The noisy mixed signal is then further denoised using a block denoising method based on the Fast Fourier Transform (FFT). The split sound signals are then reconstructed using Inverse Fast Fourier Transform (IFFT). The proposed technique is easier to remove additive noise than existing methods, and the separated sound signals have a hearing impression that is quite similar to the original signals.