Speech Enhancement Algorithm

Speech enhancement is an important technology of speech signal processing. For many years many progresses have been made in this field to find good methods for speech enhancement.Although there are various methods for speech enhancement, the principal perceptual criteria for measuring the performance of a speech enhancement system are speech quality and intelligibility. The two performance measures are not correlated and sometimes conflicting. A good speech enhancement system is a compromise between two measures according to the requirement. In this thesis, a high quality speech enhancement with auto noise estimation is presented.A complete speech enhancement system has three components: a noise estimator which is used to trace noise estimation, a noise reduction algorithm which is the crucial part in the system and a post processor which is necessary if specific design is required.Generally speaking, a noise estimation algorithm is a noise variance estimator. The accuracy of noise variance estimation will affect next process. A noise estimation based on minimum statistics is deeply studied in this thesis. The evaluation of the method is given after a simulation experiment. The speech enhancement system of this thesis will adopt the noise estimator based on minimum statistics.The comparison of threes methods, spectral subtraction, LOGSTSA-MMSE and subspace approach, is the major work in researching the noise reduction algorithm. The LOGSTSA-MMSE method is the best one of the three methods after making a comprehensive survey of increasing SNR and auditory perception. LOGSTSA-MMSE is chose in the speech enhancement system of mis thesis.The goal in post processing in the thesis is to improve quality of speech, reduce distortion of the speech and decrease the effect of "musical noise". The masking effect is used in post processing, experiment results show that speech quality is improved after post process.