Joint source-channel coding of LSP parameters for bursty channels

This work present a joint source-channel technique based on Channel Optimized Vector Quantization (COVQ) for transmission over bursty channels applied to LSP parameters coding. The bursty channel is modeled as a Finite State Channel (FSC) with two states. We call Bursty COVQ (BCOVQ) to the resulting quantization technique. The case in which channel state information is only available at the receiver is considered. The application of BCOVQ to LSP (Line Spectrum Pair) parameters coding in wideband speech coding is presented. Our experimental results show that BCOVQ maintains a good performance through all tested channel conditions, while other reference techniques (traditional COVQ, Split Vector Quantization and Split Multi-Stage Vector Quantization) are more dependent on these conditions.

[1]  Kenneth Rose,et al.  MSVQ design for packet networks with application to LSF quantization , 2000, 2000 IEEE International Conference on Acoustics, Speech, and Signal Processing. Proceedings (Cat. No.00CH37100).

[2]  Biing-Hwang Juang,et al.  Line spectrum pair (LSP) and speech data compression , 1984, ICASSP.

[3]  José L. Pérez-Córdoba,et al.  A study of joint source-channel coding of LSP parameters for wideband speech coding , 2003, 2003 IEEE International Conference on Acoustics, Speech, and Signal Processing, 2003. Proceedings. (ICASSP '03)..

[4]  Masoud Salehi,et al.  Optimal quantization for finite-state channels , 1997, IEEE Trans. Inf. Theory.

[5]  Kuldip K. Paliwal,et al.  Efficient vector quantization of LPC parameters at 24 bits/frame , 1993, IEEE Trans. Speech Audio Process..

[6]  Vinay A. Vaishampayan,et al.  Design of multiple description scalar quantizers , 1993, IEEE Trans. Inf. Theory.

[7]  William H. Tranter,et al.  The performance of Reed-Solomon codes on a bursty-noise channel , 1995, IEEE Trans. Commun..