Optimization for G .722.1 wideband speech codec based on embedded CPU

TAN Wei, MA Qi
(IC CAD Research Center, Hangzhou Dianzi University, Hangzhou 310037, China)

Abstract: Aiming at the drawback that the speech codec was mostly implemented based on DSP, but DSP had a relatively high hardware cost, a 32 bit embedded CPU platform CK510/CK520 with domestic intellectual property was applied to make the realtime implementation of the speech codec. Firstly, the G.722.1 algorithm for wideband speech codec based on modulated lapped transform technology was discussed. Then, combining the principle of the G.722.1 algorithm and the characteristics of CKCPU, some optimization strategies in C and assembly language were put forward to make the algorithm program more suitable for hardware implementation. The test results indicate that the optimized G.722.1 algorithm can rebuild the speech with high quality, while its time complexity and memory size were both effectively reduced. This research not only helps the formation of a relatively simple and cheap application solution based on single CPU for the speech codec, but also benefits the promotion of domestic CPU.
Key words: G.722.1; wideband speech coding; 32bit embedded CPU; algorithm optimization