5th International Conference on Spoken Language Processing

Sydney, Australia
November 30 - December 4, 1998

Efficient Quantization Of LSF Parameters Based on Temporal Decomposition

Sung Joo Kim, Sangho Lee, Woo Jin Han, Yung Hwan Oh

KAIST, Korea

In this paper, we present a restricted temporal decomposition method for LSF parameters. The event vectors estimated by this method preserve the ordering property of LSF parameters so that they can be quantized efficiently. Experimental results show that interpolated LSF parameters can be quantized transparently at the rate of 753bps. Also we design a LPC vocoder at 996bps as an application of the proposed method. According to a listening test, the reconstructed speech of our vocoder has reasonable quality comparing with 2400bps LPC10e.

Full Paper

Bibliographic reference.  Kim, Sung Joo / Lee, Sangho / Han, Woo Jin / Oh, Yung Hwan (1998): "Efficient quantization of LSF parameters based on temporal decomposition", In ICSLP-1998, paper 0469.