A very low bit-rate matrix quantized speech coder with Gray coding

Autor: Wee-Soon Ching, W.C. Wong, H.S. Bay
Rok vydání: 2002
Předmět:
Zdroj: Proceedings of ICSIPNN '94. International Conference on Speech, Image Processing and Neural Networks.
DOI: 10.1109/sipnn.1994.344869
Popis: This paper presents a novel very low bit-rate LPC-based speech coder operating at 385 bps. It employs matrix quantisation (MQ) of the reflection coefficients, and differential encoding and quantisation of the gain and pitch parameters. The MQ codebook comprises four sub-codebooks based on the classification of the state transitions. Instead of coding the absolute gain level, only the difference between the input gain and the average gain for the quantisation template is transmitted. For the pitch quantizer, the direction of pitch change is coded rather than the actual pitch value. All the quantisation levels are gray coded which improves robustness. The performances in term of objective measures which show the merits of employing Gray-coding into the quantized speech parameters are given. Informal listening tests show that the proposed speech coder produces good quality speech when compared to other fixed-rate speech coders that operate at 300-400 bps under Rayleigh fading mobile radio channel of different SNRs. >
Databáze: OpenAIRE