Robust vector quantization for low bit rate speech coding

Speech coding systems for mobile communication have to cope with noisy channels. In particular, vector quantization as central data reduction scheme is highly sensitive to transmission errors due to the low redundancy in the encoded data. Here we present three methods for the design of a vector quantizer with enhanced robustness against transmission errors. First the optimization of the index assignment of a LBG vector quantizer via simulated annealing is investigated. Second a neighborhood conserving vector quantizer is designed by using a topology conserving feature map. Third we discuss a method in which the error characteristic of the channel is used in the optimization of a vector quantizer as well as in the quantization of a data vector. Simulation experiments and results for a binary symmetric channel with bit error probabilities up to 10% are presented for vector quantization of speech signals and predic-tor parameters.