ASET voice coding-algorithm and implementation

The development and implementation of a medium bit-rate speech compression algorithm are reported. The algorithm, named adaptive subbands excited transform (ASET) coding, produces high-quality speech at relatively low computational complexity. Informal subjective testing of the algorithm showed no distortion or degradation in speech quality at 16 kb/s and only small degradation at 12 kb/s. In a formal mean opinion score test, 16-kb/s ASET scored 3.9/5.0 matching the score reported for standard 32-kb/s ADPCM (adaptive differential pulse code modulation). The real-time implementation is based on a single board with a single, commercially available, digital signal processing (DSP) chip. Major components of this implementation include the DSP chip, on-board data and program memories, and a dual-port memory chip. The dual-port memory is used as a buffer between I/O circuitry and the processor, allowing them to operate asynchronously. This configuration increases processing efficiency and simplifies the hardware design and operations.<
>