Musical Instrument Sound Signal Separation from Mixture using DWT and Fast ICA Based Algorithm in Noisy Environment

Abstract

The structured arrangement of sounds in musical pieces, results in the unique creation of complex acoustic mixtures. The analysis of these mixtures, with the objective of estimating the individual sounds which constitute them, is known as musical instrument sound signal separation, and has applications in audio coding, audio restoration, music production, music information retrieval and music education. In this paper, the problem of sound signal separation from the mixture in the presence of additive white Gaussian noise (AWGN) is investigated. Because of the additive noise independent component analysis (ICA) algorithm does not give a consistent estimation of separated signals. The different sound signals are first down sampled and then mixed with the help of mixing matrix found iteratively. The mixed signal is added with variable percentage of AWGN. Resulting noisy mixed signal is further denoised with block denoising algorithm, and then decomposed into wavelet coefficients with DWT. Fast ICA algorithm is further applied on these wavelet coefficients for identification of individual sound signal, and then IDWT is used to reconstruct the separated sound signals. The hearing perception of the separated sound signals is close to the original signals