Study And Application Of Speech Enhancement Algorithm Based On Adaptive Spectral Estimation

In the real environment, due to background noise speech signal is always affected by noise inevitably,which leading to the original speech signal to noise radio and intelligibility declining.With further increase of background noise,the original speech signal even will be submerged in background noise,thus unable to distinguish original speech signal. The pollution of noise will further influence the function of back-end signal processing.Speech enhancement is an effective way to solve noise pollution and plays an important position in speech signal processing.Its purpose is to extract original speech signal from noisy speech signal as much as possible,in order to restrain background noise and improve signal to noise radio and intelligibility.So many noise sources result in many speech enhancement algorithms,but no method is suitable for all kinds of noise. In this paper we studied speech enhancement depth. First we introduced the research background,history and current situation of speech enhancement and analyzed the basic theory of speech enhancement and the subjective and objective evaluation criteria.We discussed the speech enhancement algorithm based on short-time spectral subtraction,speech enhancement based on wavelet and wavelet packet transform and speech enhancement based on subspace.The accurancy of noise estimation will influence the effect of speech enhancement greatly. The current noise estimation method is divided into two categories according to wether detecing the voice activity. In the paper,we also discussed noise estimation methods in detail.Based on the study of speech enhancement based on short-time spectral,utilizing human ear speech perception characteristics,we propose adaptive spectal estimation algorithm to improve the spectral subtraction and MMSE method. In these algorithms, the noisy speech power spectrum is divided into Bark subbands, and the coefficients in each subband are calculated according to the S/N rates. The improved method suppresses noise more balanced. The simulation results show that the improved speech enhancement algorithm improved the signal to noise ratio and intelligibility with reducing the "musical noise".Cochlear implant help the deaf patients to restore hearing by translating the sound into electrical signal to stimulate the auditory nerve fibers of deafness.The technology is better in quiet environment,but deteriorate in noisy environment rapidly. In order to improve the noise immunity of CI,we applied the algorithm to the ACE synthesis,the result showed that the CI synthetic sound is more better and clearly.