Research Of Speech Enhancement Technologies In High Level Noise Environments

Speech enhancement is an important technology of speech signal processing. It aims at improving the performance of speech communication systems in noisy environments. Speech enhancement may be applied, for example, to a mobile radio communication system, a speech recognition system, or to improve the performance of aids for the hearing impaired. For many years many progresses have been made in this field in order to find good methods for speech enhancement, especially in high level noise environments.In this thesis, eight kinds of speech enhancements are firstly analyzed, and the best three of these methods are tested and compared. Test results show that many speech enhancement algorithms are plagued by an annoying artifact called "musical noise", especially in high level noise environments.A postfilter that aims to reduce musical noises in high level noise environments is then proposed. The postfilter firstly detects isolated tones in the short-time spectrum, and then discriminates musical noises from harmonics of voiced speech signal at the locations of the detected isolated tones and their neighborhoods. Finally, musical noises are suppressed and smoothed over spectral axis. The proposed postfilter is combined with the spectral subtraction based on minimum statistics. Subject tests based on ITU-T P.835and spectrogram comparisons show that with the postfilter, musical noises are efficiently removed with little distortion on speech.