Single DSP 8kbps speech CODEC
T. Miyamoto, Kazuhiro Kondo, Torazo Suzuki, Yoshiaki Asakawa, Akira Ichikawa · 2005
This paper describes the implementation of an 8kbps speech CODEC using a single DSP. First, an effective algorithm based on the TOR (Thinned-Out Residual) method has been developed by modifying the means to thin out and interpolate residuals in voiced and unvoiced frames, and to extract and modify the pitch period. Then, a compact CODEC which consists of one DSP and one MPU each has been developed using the new algorithm. This CODEC supports not only speech coding functions, but also frame synchronization functions and echo suppressing functions. The measured speech quality is quite satisfactory.