A speech coding method using LZ algorithm

摘要

Most of speech coding parameters have temporal redundancy, which might be removed. This article presents a new speech coding method using Lempel-Ziv algorithm. The proposed method was applied to quantize LP coefficients at first, and it performed better than Split-VQ, MSVQ, and MA prediction VQ in rate-distortion criterion. Then, the proposed method was also used to quantize F0 and gain, and a coder at 1.9 kbit/s was designed. The quality of the coded speech was almost compatible with the FS-MELP at 2.4 kbit/s according to the subjective tests.