Efficient product code vector quantisation using the switched split vector quantiser
In this article, we ?rst review the vector quantiser and discuss its well-known advantages over the scalar quantiser, namely the space-?lling advantage, the shape advantage, and the memory advantage. It is important to understand why vector quantisers always perform better than any other quantisation scheme for a given dimension, as this will provide the basis for our investigation on improving product code vector quantisers which, despite having much lower computational and memory requirements, result in suboptimal quantisation performance. The main focus is on improving the ef?ciency of the split vector quantiser (SVQ), in terms of computational complexity and rate-distortion performance. Though SVQ has lower computational and memory requirements than those of the unconstrained vector quantiser, the vector splitting process adds numerous constraints to the codebook, which results in suboptimal quantisation performance. Speci?cally, the reduced dimensionality affects all three vector quantiser advantages. Therefore, we investigate a new type of hybrid vector quantiser, called the switched split vector quantiser (SSVQ), that addresses the memory and shape suboptimality of SVQ, leading to better quantisation performance. In addition, the SSVQ has lower computational complexity than the SVQ, at the expense of higher memory requirements for storing the codebooks. We evaluate the performance of SSVQ in LPC parameter quantisation, used in narrowband CELP speech coders, and compare it against other quantisation schemes.
Digital Signal Processing