多带激励低速率语音压缩编码算法研究及实时实现

Research on MBE Algorithm and Implemented in Realtime with Low Bit Rate

本文以多带激励声码器为模型，采用了多种技术去降低编码速率和改善音质.我们利用动态规划算法对基音周期进行平滑，去除了声码器中常有的音调噪声.MBE算法对谱包络的量化要花费大量的比特，这里利用LPC全极点模型谱逼近MBE谱包络，并采用共振峰增强技术来补偿模型误差.对谱幅度参数的量化，采用了分裂矢量量化（SPVQ）和多级矢量量化（MSVQ）的方法.使之在2.4kbps，1.2kbps及800bps等速率上，用单片ADSP2181实时实现了较高质量的语音压缩编解码.

This paper presents a high quality variable-rate speech coding at 2.4kbps, 1.2kbps and 800bps based on the algorithm of Multi-Band Excitation (MBE). The techniques of smoothing pitch period using dynamic programming algorithm, LPC all-pole model to approximate the MBE spectral shape, formant enhancement, split vector quantizer (SPVQ), multi-stage vector quantizer (MSVQ), etc. Are adopted to improve this vocoder. It can be implemented in realtime using a single ADSP2181 fixed point DSP.The speech quality of the realtime systems reach the currrent international level. The synthesis speech is clear and intelligent,as well as intelligent being in the presence of background noise at 80dB.