Medium band speech coding using optimal scalar quantization of LSP

Samir Saoudi (1) , Jean-Marc Boucher (1) , Alain Le Guyader (2)

1 SC - Département Signal et Communications
2 FT - France Télécom, Division R&D, Lannion (France Télécom)

The most popular LPC parameters used for speech Linear Predictive Coding at medium or low bit rate are the reflection coefficients and the Line Spectrum Pairs. In this paper, comparisons of the quantization performance of these two types of coefficients will be given. Several algorithms are studied for the quantization of the LSP parameters. The performance of the quantization methods of the LSP parameters is studied for a different total number of bits per frame. The Fox-Makhoul procedure is used to derive the optimal bit allocation to achieve the best quantization performance for both LPC parameters (Line Spectrum Pairs (LSP) and Log Area Ratio (LAR)). The quantization tables of the LSP and the LAR parameters are introduced in a 8 kbits/s CELP coder for comparison.

Domaines: Traitement du signal et de l'image [eess.SP]

Samir Saoudi, Jean-Marc Boucher, Alain Le Guyader. Medium band speech coding using optimal scalar quantization of LSP. EUROSPEECH '91 : european conference on speech communication and technology, Sep 1991, Genova, Italy. pp.905 - 908. ⟨hal-01864884⟩