Research On Speech Enhancement Based On MMSE-LSA

Speech Enhancement is an important branch of speech signal processing. It is of great value not only in daily life but also in military and speech signal processing technology. Speech enhancement is relevant to speech processing theory, human beings' auditory properties, phonetics and other fields.Firstly, Speech enhancement based on MMSE-LSA is researched. Some speech enhancement algorithms are introduced and the speech enhancement based on MMSE-LSA is improved, the noises are effectively suppressed and better subjective auditory quality is gained. The gain function of MMSE-LSA relies on the prior SNR and posterior SNR. To compute prior SNR and posterior SNR, the estimation of noise is needed. The method that updates the noise estimation continuously using smoothing parameters is proposed and it can better adapt to the non-stationary environment. In order to suppress the residual noise effectively while reduce the chance of distortion of speech, the noise masking property is combined with the speech enhancement and the results show that it can gain better performance.Secondly, the pre-processing of noisy speech is primarily studied, the method that noisy speech can be pre-processed by linear prediction and mathematical morphology before being enhanced. The MMSE-LSA has better outcomes especially for white noise, for the noisy speech corrupted by other noises, the pre-processing procedure can be adopted to ensure the performance of the speech enhancement.