A comparison of LSF quantization techniques for 3G mobile communications
R. McCarthy · 2004
This paper uses the 3rd Generation (3G) adaptive multi rate (AMR) speech coder to evaluate the performance of four weighting functions (WF) for use in the quantisation of line spectral frequency (LSF) values. The LSF are a transformation of the speech synthesis filter coefficients. Each WF performance is analysed in conjunction with both split vector and split matrix LSF quantisation techniques. A spectral distortion (SD) measure using the quantised and unquantised LSF is generated to determine the best WF for each quantiser. The results show that the adaptive WF (AWF) performs best in the split vector case and the 3G AMR WF performs best in the split matrix case.