Optimizing the synthesis filter bank in audio coding for minimum distortion using a frequency weighted psychoacoustic criterion

This paper shows the improvement brought in filter-bank based audio coding schemes by relaxing the constraints of perfect reconstruction or maximum frequency selectivity on the synthesis filter bank: here, the coding scheme is optimized according to a frequency weighted psychoacoustic criterion. This improvement is then evaluated by means of rate/distortion curves (the weighted criterion being the distortion measure). This approach does not take frequency masking effects into account but is compatible with their use.