CS reconstruction of the speech and musical signals

TL;DR

This paper explores the application of compressive sensing to speech and musical signals, analyzing sparsity in Fourier and cosine domains, and compares reconstruction performance with varying sample sizes.

Contribution

It demonstrates the effectiveness of compressive sensing for both speech and musical signals, highlighting the advantages of the cosine domain for signal reconstruction.

Findings

CS successfully reconstructs both signal types

Speech signals require more samples for accurate reconstruction

Cosine transform domain yields better results with fewer samples

Abstract

The application of Compressive sensing approach to the speech and musical signals is considered in this paper. Compressive sensing (CS) is a new approach to the signal sampling that allows signal reconstruction from a small set of randomly acquired samples. This method is developed for the signals that exhibit the sparsity in a certain domain. Here we have observed two sparsity domains: discrete Fourier and discrete cosine transform domain. Furthermore, two different types of audio signals are analyzed in terms of sparsity and CS performance - musical and speech signals. Comparative analysis of the CS reconstruction using different number of signal samples is performed in the two domains of sparsity. It is shown that the CS can be successfully applied to both, musical and speech signals, but the speech signals are more demanding in terms of the number of observations. Also, our results show that discrete cosine transform domain allows better reconstruction using lower number of observations, compared to the Fourier transform domain, for both types of signals.