Intraframe quantization of line spectrum pairs for narrowband speech coding in packet networks

The standard ITU-T G.729 coder uses an interframe quantization of the line spectrum frequencies (LSF) parameters which causes error propagation to the next frames. We propose the use of intraframe quantization schemes to overcome this problem. We give a performance comparison between intraframe and interframe quantization methods of LSP parameters for ITU-T G.729. Simulations results show that our intraframe quantization method achieve smaller average spectral distortions than interframe quantization embedded in the ITU-T G.729. Average spectral distortions (Avg. SD), perceptual evaluation of speech quality (PESQ) and Enhanced modified bark spectral distortion (EMBSD) tests confirm that the proposed methods are superior to the interframe algorithm embedded in the ITU-T G.729.