Low-complexity encoding of speech lsf parameters using multistage tree-structured vector quantization: Application to the MELP coder

摘要

A fast search procedure to reduce the search complexity required to locate the codevectors during the encoding process in multistage tree-structured vector quantization (MTVQ) is proposed. Quantization of line spectral frequency (LSF) parameters at different rates is used to provide experimental results, which are compared to the multipath sequential search or M-L search (MSS) and the multipath sequential tree-assisted search (MSTS). The Federal standard MELP coder is modified by replacing the original LSF quantizer with an MTVQ using the proposed fast search procedure, and an evaluation of the produced speech quality is given.