Research And Application Of Speech Enhancement Arithmetic

As a natural and convenient medium of controlling and communication,the way of using speech signals to realize man-machine conversation directly has been applied to each practical field widely. And it's validity has been testified. At the same time, as the most popular and directly expression manner of information, speech signals also have wide application foreground. However the key technology of speech signal processing at practical circumstances is anti-noise,because noise reduction is necessary for low cluster of speech recognition.Usually Speech Enhancement is used as the module of pretreatment or front end production,and it is also an important pretreatment technology of speech coding, speech recognition and speechsynthesis. With the development of speech technology all kinds of speech production systems are exposed to the problem of performance improvements. Speech Enhancement is one key technology all this. Speech Enhancement is that using some signals processing technologys to pick up useful speech signals from noise,to raise SNR, enhance intelligibility and make man hear unwearied. High Speed Digital Signal Processor(DSP) developed in recent years is specially designed for algorithms of digital signal processing. Nowadays, it is widely used in many fields such as communications,speech processing,video processing,etc.. This essay focuses on the research and real-time implementation of speech enhancement algorithms on speech processing platform.Firstly the paper introduced some classic speech enhancement algorithms including Noise Cacellation,Spectral Suppression and Wiener filtering,and compared and analysized their excellence and weakness. Author made some improvements on these algorithms and introduced a modified spectral suppression algorithm based on extended spectral substraction. This algorithm is easy to be implemented real-time on the hardware system and could satisfy the demands of real time speech enhancement and validated its effect through simulation.The paper introduced the hardware and software system based on the TMS320VC55x, the chips and system,the initialization software modules,and the implementation of the algorithm. The primary components of this system are VC5509 and codec TLV320AIC23B. The test showed that this system could have a good effect on the reduction of the background noise and get a better understanding of the speech so as to achieve the goal of speech enhancement.