基于低信噪比环境下音频净化技术的分析-analysis of audio purification technology based on low signal-to-noise ratio environment.docxVIP

The Audio Purification Technology Research Based on Low SNRAbstractThe human voice is the quickest, most effective and most important way to exchange information. With the rapid development of audio purification technology, The traditional approach has been unable to achieve the desired effect in low SNR environment, When the noise seriously affect the performance of the system, so, how to improve the purification effect is necessary to be resolved at this stage. The purpose of this thesis is that from the sound signal with noise as possible to extract pure sound signal, and the background noise will be suppression. So, the original voice can understand easy. The audio purification technology research based on low SNR environment mainly used for listening area (catch the criminals). Meanwhile, the technology involves a wide range of applications, Such as hearing aids, electrical cochlea, human-computer interaction systems and mobile voice communications and other aspects, they has high research value. In this context, the main work of this paper is as follows:In this thesis, a zero rate on short-term average endpoint detection method be researched. The method is separation using sound signals and noise signals, voice signals and accurately determine the starting point of the noise signal, the completion of the preprocessing of the audio purification system;Based on wavelet transform MFCC feature extraction method researched in this thesis. This method represents an organic sound signal static characteristics and dynamic features, MFCC can simulate human hearing characteristics, wavelet transform can be a good deal signals a change process, completed the sound signal to extract effective characteristic parameters;This paper presents a neural network based on quantum hearing audio purification method. This method implements the training process the sound signal pattern matching, using quantum neural network nonlinear mapping and self-learning ability to optimize there