Research On Speech Coding Model Of EV-CELP

Speech coding technology in the communication system plays an important role, with the rapid development of communication technology, speech coding technology becoming more demanding, so they need a speech coder with better speech quality at lower encoding rate.In this paper,4.8kb/s CELP speech coding technology is studied, in order to reduce the coding rate and ensure voice quality, focusing on the voice activity detection and speech enhancement techniques. A new speech enhancement algorithm based on wavelet packet and teager energy operator is presented。Noise reduction of the algorithm can be better than the fixed thresheolds method in speech. The paper analyzes the classic speech endpoint detection based on shor-time energy and short-time zero-crossing rate feature. Accuracy of the algorithm is good in high SNR environment, and worsen in low SNR environment.A new endpoint detection algorithm is presented in low SNR environment, which can be adaptive threshold and has better robust capability. The improved algorithm is fit for actual requirement. In this paper, the speech enhancement algorithm and the new endpoint detection algorithm are applied to CELP encoder, and EV-CELP model is presented. Simulations in MATLAB demonstrates EV-CELP model, and performance testing is made. The results prove that it is both low code rate and good rebuild speech.EV-CELP model is high-quality low bit rate speech coding algorithm. Its quality reach require of communication, which can be applied communication system.