Over-sampling for accurate masking thershold calculation wavelet packet audio coders

Many existing audio coders use a critically sampled discrete wavelet transform (DWT) for the decomposition of audio signals. While the aliasing present in the wavelet coefficients is cancelled in the decoder, these coders normally perform calculation of the simultaneous masking threshold directly on these aliased coefficients. This paper uses over-sampling in the wavelet packet decomposition in order to provide alias-free coefficients for accurate simultaneous masking threshold calculation. The proposed technique is compared with masking threshold calculation based upon the FFT and critically-sampled wavelet coefficients, and the results show that a bit rate saving of up to 16 kbit/s can be achieved using over-sampling