Research On G.729 Speech Codec Algorithm And Implementation On DSP

The fundamental of speech coding is the fact that how to get better quality of reconstructed speech and ensure the less time delay and proper operational complexity at the given coding rate. In 1996, ITU-T produced G.729 standard, it's a 8kb/s conjugate-structure algebraic code excited linear prediction speech coding project, which is commonly abbreviated CS-ACELP. The characteristics of this project are as follows: the analytical window adopts hybrid window; linear spectral pairs uses two-stage vector quantization; the codebook search using sub-frame as unit can be classified as adaptive codebook search and algebraic codebook search; pitch analysis uses the joint of open-loop analysis and adaptive codebook search, reducing the operational quantity and quantized bit numbers of pitch, improving the accuracy of pitch prediction; easy algebraic code without storing codebook, clearly reconstructed audio quality. G.729 standard has been widely used on VOIP gateway,IP phone because of its nicer synthesized speech quality and appropriate complexity and less time delay. This article first introduces the background knowledge of low-rate speech coding, and specifies G.729 standard in detail, analyzes the important technologies in theory: linear prediction, vector quantization, codebook search and post-filter. Then, it introduces the characteristics of low power loss and high-performance 16-bit fixed-point digital signal processing chip TMS320VC5410, optimizes the larger operational quantity modules of the source code. Finally, it implements G.729 standard algorithm on DSP. The simulated result shows the synthesized speech has better naturalness and intelligibility.