Most of the current audio codecs decompose the input signal into spectral components. These spectral components can then be coded more efficiently than the full bandwidth input signal. The quantization is controlled by a psychoacoustic model so that the audibility of the resulting distortions is kept minimal. Subsequent entropy coding aims at a minimization of the resulting bit rate.
The individual components are to be optimized in order to increase the coding efficiency of the overall system.
Focus is given on the following topics: