Speech reconstruction based on compressed sensing theory using smoothed L0 algorithm

At present, speech signals have good sparsity in frequency domain and discrete cosine transformation (DCT) domain etc. Therefore it can be researched based on compressed sensing (CS). CS compresses signals which are sparse or compressible at the sending end, and then reconstruct them at the receiving end. This paper proposes the compressed speech reconstruction method based smoothed L0 (SL0) algorithm. Simulation results demonstrate that the SL0 algorithm can obtain a better performance than the traditional orthogonal matching pursuit (OMP) method in reconstruction of speech signals.