engleski

Variable bit-rate LSF coding using Wavelet sub-band decomposition

Variable bit-rate Wavelet sub-band coding of the Line Spectrum Frequencies (LSF) is presented in this paper. The evolution of the speech power spectrum envelope (PSE) described by the LSF vector sequence is decomposed into several parts, e.g.: very rapid transitions, medium speed transitions and slowly evolving part. All these parts are coded independently, according to the transformed data dependent Weighted Euclidean Distance (WED) measure between the original and the quantized LSF vector. Forward adaptive quantization is realized by employing the dynamic weighting path into the non-uniform filter bank realization of the Discrete Wavelet Transform (DWT). Vectors of the sub-band signals are classified using the nearest neighbor rule, based on the square root weighted sub-band signal magnitudes. Independent vector quantizers are designed for each of the classes and for each of the sub-band signals. Thus the required bit-rate for the speech PSE coding is adapted according to the changing dynamics of the speech signal.

speech processing; spectral quantization; inter-frame sub-band coding; adaptive quantization; line spectrum frequencies

nije evidentirano