An Efficient LSF Parameters Quantizer With Low Storage

To reduce the computational complexity and storage of the line spectrum frequency(LSF) parameters quantizer a predictive multi-stage split vector quantizer is designed to quantize LSF parameters by using the correlation of LSF parameters between frames and inter frames. The LSF residual vector is quantized by 2-stage vector qutization (VQ) and error vector is divided into 4-D and 6-D subvectors in the second stage. This vector quantizer achieves an average spectral distortion of less than 1 dB at 20 bits/frame with low computational complexity and storage, where the mulit-stage VQ simultaneous joint design algorithm is used to train the codebooks and the M-Ltree search algorithm is applied to select codes.