Research On Single Channel Adaptive Speech Enhancement

Voice is a convenient form for humans to obtain information,but voice calls are often interfered by various noises from the environment.This not only causes the quality of voice calls to be reduced,but affects the normal system when the noise intensity reaches a certain level.In actual speech signal processing,it is often desired to obtain high-quality speech for efficient transmission.Therefore,the research on the anti-noise of speech has corresponding theoretical research value and practical application significance.In modern intercom systems,such as the call systems of helicopters and tanks in the military field,the quality of voice calls is seriously affected by the interference of harsh environmental noise.In order to improve speech quality,the thesis carried out adaptive enhancement research on single channel speech,the main work is as follows:Firstly,the basic knowledge of speech signal processing is introduced,which includes speech signal characteristics,human ear perception characteristics,speech ignal mathematical models,speech signal preprocessing,and performance evaluation of speech enhancement algorithms.Analyzed and discussed three classic single channel speech enhancement algorithms:spectral subtraction algorithm,Wiener filtering algorithm and minimum mean square error estimation,under the background of white noise and babble noise,completed the simulation of three algorithms based on MATLAB software,compared and analyzed the performance of the three algorithms and gave the application scope of these three algorithms and their respective advantages and disadvantages.Secondly,the spectral subtraction algorithm and the Wiener filtering algorithm are only applicable to the defects of stationary noise and the minimum mean square error method is not ideal for non-stationary noise background,in which the reduction effect is not ideal and the algorithm implementation complexity is high,and the difficulty of transplantation is difficult.Based on the canceller,a single channel adaptive algorithm based on an adaptive noise canceller under delay conditions was constructed and analyzed.The implementation principle of the algorithm's noise reduction is emphasized and the performance simulation based on MATLAB under white noise and babble background noise was carried out,and compared with the three classic algorithms.The simulation results showed that the single channel adaptive algorithm has better speech adaptability and noise reduction performance under low SNR conditions.Finally,the application transplantation of single channel adaptive algorithm was realized.Based on the KEIL IDE integrated development environment and the STM32F407 hardware and software platform,the algorithm code was transplanted and developed.The real-time speech noise reduction algorithm based on the single-channel adaptive algorithm was tested.