Two-stage bit allocation algorithm for stereo audio coder

A new bit allocation algorithm is proposed and applied in a stereo transform coder. This algorithm is a two-stage approach based on the minimisation of perceptual distortion measured by artificial ears. At the first stage, the use of minimum frequency-weighted reconstruction error is exploited, in which the weighting factors are related to the masking threshold and the absolute hearing threshold. The bit assignment obtained at the first stage is fine tuned by a greedy algorithm of perceptual distortion versus bit rate at the second stage. A joint stereo perceptual transform coder is implemented and incorporated with this two-stage bit allocation algorithm. The listening tests show that this coder performs better than MPEG layer 2 at bit rates below 128 kbps and has the comparative quality of MPEC layer 2 at higher bit rates