An improved audio encoding architecture based on 16-bit fixed-point DSP

In this paper, we present a novel MPEG audio layer III encoder architecture. Methods to reduce the heavy computational load are proposed, including a new psycho-acoustic model based on MDCT, a fast bit allocation strategy without many iterations and a suitable algorithm of non-linear functions. Adopting these methods, we implemented a real-time encoder on a 50 MIPS 16 bit fixed-point DSP, which can produce an ISO/MPEG compliant bit stream. Results show that this architecture can achieve near-CD quality at a bit rate of 128 kbps for dual channels with greatly reduced computational load. It is suggested that this improved audio encoding architecture can better satisfy the requirements of the consumer market.