Recursive architecture using pipeline stall for low complexity MPEG-4 ALS encoder

Abstract

An MPEG-4 audio lossless coding (ALS) encoder consists of a short-term predictor, a joint coder, a long-term predictor, and an entropy coder for audio signal compression. Especially, the short-term predictor is the highest complexity in MPEG-4 ALS encoder implementation because linear prediction coefficients (LPCs) for the short-term predictor are obtained from the Levinson-Durbin recursion algorithm that uses correlations of the input signals. LPCs are used for the finite impulse response (FIR) filter of the short-term predictor and its maximum coefficient order is up to 1023. Proposed is an area-efficient short-term predictor through the recursive architecture using pipeline stall and the modified radix-2 FIR algorithm is proposed. The proposed shot-term predictor has a reduction of complexity by about 95% compared to the radix-2 1024 tap FIR filter.