Speech Enhancement Based On Hilbert-Huang Transform

Presently, speech communication gains widely application in a variety of fields, but in the process of speech communication, the speech signals are often corrupted unavoidably by ambient noise, electrical noise in transmission media, electrical noise in communications facility and the speech of other speakers. These all result in a noisy speech signal instead of the original speech signal that the receiver should get. Therefore, the research on the technology of speech enhancement plays an important role in the quality of speech communication.Speech enhancement can filter the noisy speech signal to improve the quality of it, and make the speech more clearly and comprehensible so as to be accepted easily or improve the performance of the speech processing system.The technology of speech enhancement that based on the empirical mode decomposition (EMD) is researched. We firstly introduce some speech enhancement algorithms and the Hilbert-Huang transforms theory (HHT). Then we explained explicitly the performance of the application of the EMD to speech signal, and we researched the relations of IMF components in energy. The way in which the EMD decomposes is similar to the way that the wavelet transform does. That is to say, the decomposition result is distributed from high frequencies to low frequencies, and the noise concentrates mainly on the anterior layers. Then we proposed a speech enhancement method that based on EMD and Linear Prediction, and we also proposed a method that based on EMD and Spectral Subtraction. The experiment shows that the method could gain excellent performance, especially for the white noise.