Title :
Novel dynamic bit allocation method for LSP quantization
Author :
Hui, Fang ; McLoughlin, Ian
Author_Institution :
Sch. of Comput. Eng., Nanyang Technol. Univ., Singapore
Abstract :
The line spectral pair (LSP) conveys vocal tract information for reconstruction of speech in an analysis-synthesis speech coding system. Quantization of LSP parameters is necessary to reduce bit-rate, but causes unavoidable degradation in quality. This paper proposes an adaptive quantization method to reduce bit-rate whilst maintaining reasonable speech quality. The method exploits the relationship between LSP position and speech features
Keywords :
spectral analysis; speech coding; LSP parameters; LSP quantization; adaptive quantization method; analysis-synthesis speech coding system; bit-rate reduction; dynamic bit allocation method; line spectral pair; speech quality; speech reconstruction; vocal tract information; Bit rate; Degradation; Frequency domain analysis; Information analysis; Linear predictive coding; Quantization; Robustness; Spectral analysis; Speech analysis; Speech coding;
Conference_Titel :
TENCON 2000. Proceedings
Conference_Location :
Kuala Lumpur
Print_ISBN :
0-7803-6355-8
DOI :
10.1109/TENCON.2000.893714
