The perceptual impact of different quantization schemes in G.719

Abstract: In this thesis, three kinds of quantization schemes, Fast Lattice Vector Quantization (FLVQ), Pyramidal Vector Quantization (PVQ) and Scalar Quantization (SQ) are studied in the framework of audio codec G.719. FLVQ is composed of an RE8 -based low-rate lattice vector quantizer and a D8 -based high-rate lattice vector quantizer. PVQ uses pyramidal points in multi-dimensional space and is very suitable for the compression of Laplacian-like sources generated from transform. SQ scheme applies a combination of uniform SQ and entropy coding. Subjective tests of these three versions of audio codecs show that FLVQ and PVQ versions of audio codecs are both better than SQ version for music signals and SQ version of audio codec performs well on speech signals, especially for male speakers.