CS-ACELP Speech Coding Simulink Modeling, Verification, and Optimized DSP Implementation on DSK 6713

The emerging of real-time telecommunication systems over low bandwidth channels enforces constraints on the transmitted data rate and it also needs an optimum speech quality at received destination, so many standards of hybrid speech coding techniques are being developed. Conjugate Structure Algebraic Code Excited Linear Prediction (CS-ACELP) hybrid speech coding was being utilized and optimized in the current work, Firstly, a CS-ACELP frame analysis with MATLAB SIMULINK model was implemented and its corresponding DSP based C++ code was generated too. Then additional optimizations were done, then, a CS-ACELP encoder/decoder with International telecommunication union (ITU) MATLAB CS-ACELP code was developed, and then MEX tool was used to generate C++ files that could compile and run it into desktop PCs. This desktop application is helpful in chat applications that require low bandwidth over the Internet and low connection speed. The performance of both DSP and Desktop implementations are evaluated using different quality evaluation tests. These tests are segmented signal-to-noise ratio (seg-SNR), log-likelihood ratio (LLR), mean opinion score (MOS) and praat software, The results showed that the optimizations done in the MATLAB Simulink work decreased the processing time by 8.564 us and reduced the used memory by 8% with acceptable speech quality.