Efficient Two-step Spectrum Quantization Methods for Low Rate Speech Coding
HR Sadegh Mohammadi, W.H. Holmes · International Symposium on Information Theory and its Applications · 1994
Line Spectral Frequencies (LSFs) are often used as parameters to represent the vocal tract filter in speech coders using linear prediction. We propose two new methods for the quantization of the LSPs, namely Combined Scalar-Vector Quantization (CSVQ) and Fine-Coarse Split Vector quantization (FCSVQ). Both of these methods are based on a two-step vector quantization scheme. The paper explains the principles of these methods, including training of the associated codebooks. It is shown that they can be implemented efficiently with negligible computational overhead compared to simple scalar quantization. Satisfactory performance of the new methods is verified through experimental tests using computer simulation.