Vector quantization of speech spectral parameters using statistics of dynamic features