Perceptual audio coders use an estimated masked threshold for the determination
of the maximum permissible just-inaudible noise level introduced by quantization.
This estimate is derived from a psychoacoustic model mimicking the
psychoacoustics of masking. Current applications use a uniform spectral
decomposition as first stage of that model to approximate the frequency selectivity
of the human auditory system. The availability of efficient implementations led
to a virtually exclusive use of uniform decompositions in audio coding.
However, the equal filter properties of the uniform sub-bands are not in line with
the nonuniform auditory filters. This paper presents a psychoacoustic model
based on an efficient nonuniform cochlear filter bank with a simplified
less complex post-processing for estimating the masked threshold. Application
results in audio coding show a significantly better performance in terms of bit
rate and/or quality of the new model in comparison with other state-of-the-art
models with a uniform spectral decomposition.
