An improved psycho-acoustic model for MPEG 1 using a Morlet Cambridge wavelet

Many existing MP3 encoder employ an FFT-based transform to derive a spectral decomposition of the audio signal into uniform subbands with equal bandwidths. The nonuniform spectral resolution of the auditory system is taken into account. The initial audio signal processing within the psycho-acoustic model consists of a spectral decomposition to account for the frequency selectivity of the auditory system. However, the auditory system performs a nonuniform spectral decomposition of the acoustic signal in the cochlea. This paper presents a psycho-acoustic model based on an efficient nonuniform cochlear filter bank following the Cambridge model. Results of the proposed psycho-acoustic model applied to audio coding show better performance in terms of compression ratio and sound quality in comparison with the classical FFT model.