Fast LSP Calculation And QuantizationWith Application To The CELP FS1016 Speech Coder

Line Spectrum Pair (LSP) representation is used for spectral quantization in the CELP FS1016speech coder, where the LSPs are first calculated, and then quantized using 34-bit non-uniformscalar quantization. In the algorithm proposed in this paper, computational complexity isdecreased by searching the zero-crossings on the grid formed by the values of the quantizationtables. As the actual LSPs are not calculated, two criteria to select the “closest” quantized LSPsare proposed. These criteria take into account the interaction between successive LSPs. Theefficiency and reliability of the proposed algorithm are improved using the interlacing property ofthe LSPs and knowledge of the direction of the sign-change at every zero-crossing. The proposedalgorithm is compared with the existing Kabal’s algorithm (followed by quantization), showingsimilar quantization performance. The computational complexity on a fixed-point DSP56001implementation is reduced by 66 %, using the proposed algorithm.