Adaptive Speech Enhancement Based On Discrete Cosine Transform In High Noise Environment

Presently, speech communication gains more and more widely application in variety of fields, meanwhile the research of speech enhancement technology plays very important role in improving the quality of speech communication.The methods of speech enhancement based on empirical mode decomposition (EMD) and discrete cosine transform (DCT) are researched. We introduce some speech enhancement algorithms and Hilbert-Huang transforms theory (HHT). The result of the estimation of noise was accurate using DCT method. Classical method based discrete Fourier transform (DFT) is not correct to view the phase as unimportant, but the reconstructed speech quality will be affected for high levels of additive noise. DCT coefficients are real and can be considered to have a binary phase value. The phase will depend only on the sign of the coefficient. Therefore if strong speech energy is present in a particular coefficient, it is unlikely that the phase will be corrupted. It is therefore likely that DCT would perform better than DFT. The experiment shows that the method has excellent performance, the performance is better for white noise.Then the performance of the application of the EMD to speech is explained explicitly, and the relations of IMF components which contrast to the result of the wavelet transform is researched. EMD similars the wavelet transforms, the decomposition result is from high frequency to the low frequency distribution, namely the noise mainly concentrates in first several. Therefore,we may use the suitable threshold value function, removes the noise in the decomposition result, retains the coefficient of the signal, again carries on the resconstruction by the surplus coefficient, mays then to restore the signal, achieves the goal of denoise. This article uses the probability threshold value method. A speech enhancement method based EMD and DCT is proposed.