Speech enhancement based on Emd and compressed sensing

Compressed sensing, as a novel framework, needs a small amount of characteristic datas to reconstruct the original datas and breakthroughs the limitation of the Nyquist sampling law, which effectively relieves the pressure of dealing with a large amount of datas in the practical application. But the noise reduction performance in traditional compressed sensing is very poor, in order to improve the situation, this paper proposes a new system that combines compressed sensing techniques with empirical mode deposition signal analysis method, which possesses both the advantages of the compressed sensing and empirical mode deposition. Experimental results show that the system not only can reconstruct the speech signal well, but also plays good noise reduction effect.