With the development of the science and the computer's popularization, people has more demands on the means of communication with the computer, which hastens the development of speech recognition and make it a very important research direction in the field of speech signal processing. At present, the correct recognition ratio of speech recognition system is very high in the lab environments, and there are some products on speech recognition. But on the effect of the noise in surroundings, the correct recognition ratio is much lower than in the lab. So the noise is the biggest obstacle in extensively practical application. It also becomes more important to research speech recognition in noisy environment.There are three kinds of de-noising processing technology, including speech enhancement, speech parameters pick-up and speech model compensation. Because of the difference of the kind of environment noise and the extent of being disturbed, it is hard to get satisfied performance using a single method usually. With the development of the de-noising speech recognition technology, how to combine rationally these methods in practical speech recognition system and make it get more performance in different noisy environment is a important research direction in de-noising speech recognition domain. In addition, there are some new technologies which combine some technologies of other domain to enhance performance.Wavelet transform, which developed in late 1980's, is a new kind of time frequency analytic method. It has good localization characteristic in both time and frequency domain; and has been applied for signal de-noising fields. In this thesis, focusing on wavelet transform theory, we analyze its characteristic. In addition, some de-noising methods based on wavelet transform is compared, the wavelet threshold method is studied in the article.In this thesis, two of de-noising technologies are combined: speech enhancement and speech parameters pick-up, and wavelet transform technology is introduced, and a method is proposed which is called combined de-noising method based on wavelet transform. Its principle is that speech signal with noise is enhanced by the wavelet threshold method to get the first de-noise in speech recognition process; then the speech parameters pick-up technology is combined with wavelet transform to get the second de-noise in speech recognition process. In this paper, a emulation experiment is used, the de-noising effect between this de-noising method and traditional de-noising method is contrasted, and it is showed that this de-noising method is effective by analyzing the result of the experiment. |