High quality MPEG-audio layer III algorithm for a 16-bit DSP

This paper presents a high quality MPEG-1/2 layer-III decoding algorithm, which is optimized for a 16-bit fixed-point arithmetic. Motorola DSP56600 16-bit DSP family processor is chosen as a target processor, and quantization noise due to fixed-point arithmetic is minimized for each decoding routine via C language-based fixed-point simulations. To maintain high audio quality with limited data precision, finite word length effects of each routine are evaluated. Also, double-precision arithmetic is tried for each routine and their effects are analyzed. Furthermore, to compromise the audio quality with the system complexity, the evaluation is extended to several fast algorithms for IMDCT and sub-band synthesis. To measure the quality of decoded output, ISO/IEC 13818-4 compliance test sequence is applied to the proposed algorithm. This test showed that the proposed decoding algorithms ensured full compliance with ISO/IEC 13818-3 audio decoder