Research And Implementation On Adaptive Multi-Rate Speech Codec

AMR speech codec was developed by 3gpp as the mandatory speech codec to the 3G system. AMR source coding bit rate is dynamically controlled by the radio network according to the channel conditions without degrading speech quality. In AMR, the most complex codebook search module is becoming the choke point. How to optimize the algorithm to reduce the computing complexity and depress the delay to achieve better communication quality is becoming the hotspot.After the research on AMR encoder and decoder in details, a fast search method of algebraic codebook in CELP coders is proposed. Sequence of codebook search is reordered according to the criterion of mean squared weighted error between target vector and filtered adaptive codebook vector, and the algebraic codebook is searched until a predetermined threshold is satisfied. There is a slight degradation about 0.1db of speech quality compared to the original one.Based on the DSP of TI's TMS320C64X series, the methods mentioned above are used to optimize and a scheme focus on digital media processors is designed. The objective evaluation based on ITU standard speech test sequences show that the optimized signal codec take 5% CPU resource, 80% less than the original one. The subjective evaluation test show that the recovered speech preserves the characteristic of speaker, the naturalness and the comprehensibility of speech being synthesized without noise is good. The real time capability and the efficiency could satisfy mostly multi-media applications.