Research And Implementation Of Speech Enhancement Algorithm In VoIP

With the rapid development of VoIP technology in recent years, Internet voice communications applications become more popular. The voice quality of IP phone is an important factor restricting its development. Many new technologies have been applied to improve the voice quality and meet the needs of users. In order to ensure the VoIP system will perform as intended and provide better voice transmission quality in a variety of environments, especially in noisy environment, it is necessary to develop a speech enhancement algorithm that can be used in VoIP system.Speech enhancement algorithm has become an important branch of the speech signal processing field. The algorithm mainly based on the different characteristics of voice and noise, estimates the background noise from the polluted voice. This paper studies a speech enhancement algorithm, based on calculating a priori and a posteriori signal to noise ratio, and then obtained a gain value that act on the noisy speech. Finally, the enhanced speech will be clearer and more comfortable by eliminating noise.In this paper, firstly, some kind of existing speech enhancement algorithms will be introduced, especially the algorithm based on short-time spectral estimation. By comparing the characteristics of several algorithms, this paper mainly studies STSA-MMSE algorithm, which is suitable for VoIP system. The implementation in this thesis contains the conversion from the floating point to fixed-point in C, and the translation of fixed-point C to DSP assembly codes with optimization. Then debug and simulate the optimized codes. The elimination effect of white noise as well as color noise will be proved. After the implementation on DSP platform, the improvement of speech quality is evaluated by P.862.