RESTRUCTURED AUDIO ENCODER FOR IMPROVED COMPUTATIONAL EFFICIENC

摘要

In an audio encoder such as AAC, the input PCM samples are sent in parallel to anMDCT and a psychoacoustic model that performs time-to-frequency decomposition onceagain with a FFT. To avoid this redundancy, we restructure the encoder to take theoutput of the MDCT as the input of the psychoacoustic model. Preliminary tests showthis idea is feasible.