A wavelet auditory model and data compression (Q1317200)

The authors discuss the Wavelet Auditory Model (WAM), which models the processing in the cochlea as a wavelet transform of the stimulus signal. The model involves processing the wavelet-transformed signal by scaling derivatives in each variable, finding ways of restricting the scale further (where are these derivatives zero?) and performing irregular sampling (frame properties) and compression. The authors give an iterative algorithm for reconstruction and present some numerical examples showing the effectiveness of this algorithm for speech compression. The scale restrictions arise through the consideration of nonlinear filters.