Multi-frame Coding of LSF Parameters Using Block-Constrained Trellis Coded Vector Quantization

In this paper, the predictive block-constrained trellis coded vector quantization (BC-TCVQ) schemes are developed for quantization of multiple frames line spectral frequency (LSF) parameters. The consecutive LSF frames are interleaved to subvectors for trellis modeling. The predictive BC-TCVQ systems are then designed to encode multi-frame LSF parameters. The performance evaluation of proposed schemes is compared with the single-frame LSF encoding methods using multi-stage vector quantization (MSVQ) and predictive block-constrained trellis coded quantization (BC-TCQ), demonstrating significant reduction of bit rate for transparent coding. The developed multi-frame LSF quantization schemes show satisfactory performance even at very low encoding rate and thus can be efficiently applied to the speech coders with moderate delay.