Comparing scalar and lattice based encoding of wavelet coefficients in sinusoidal plus wavelet packet coding

We have recently proposed a combined sinusoidal and Wavelet Packet Transform (WPT) codec for joint speech and audio coding. In this paper, we compare the performance of scalar and Z_n lattice based encoding of the WPT coefficients. The quantization of the WPT coefficients are determined by a psychoacoustic masking model, and it is identical in both cases. The audio quality remains the same for the scalar and the Z_n lattice based encoding, as only the encoding of the quantized WPT coefficients is changed. The mean bit rate of the coder (depending on the encoded signal) was reduced from 62-32 kbps to 54-30kbpsby applying the Z_n lattice based coding. Demonstration sound files are available at www-sc.enst-bretagne.fr/^~ fek/demo/.