Audio Signal Denoising Based on Laplacian Filter and Sparse Signal Reconstruction

Abstract

Impulsive disturbance commonly appears in audio signals. During the last few decades, a number of denoising approaches has been proposed for the removal of this particular type of noise. An advanced class of denoising algorithms emerged from the recent compressive sensing (CS) paradigm. Sparsity or high concentration of audio signals in some specific transformation domains, such as, for example, the discrete cosine transform (DCT) domain, can be engaged in procedures for the detection of corrupted samples. Since in the case of impulsive noise only a subset of samples is highly corrupted, upon detection of their positions, these disturbed samples can be further considered as unavailable, and reconstructed using sophisticated CS procedures. In this paper, we investigate the possibility to apply Laplacian filter in conjunction with the compressive sensing reconstruction, in the removal of impulsive disturbance from audio signals.