Multi-coder vector quantizer for transparent coding of wideband speech ISF parameters

Modern low bit-rate speech coders require efficient coding of the linear predictive coding (LPC) coefficients. Immittance Spectral Frequencies (ISF) and Line Spectral Frequencies (LSF) are currently the most efficient transmission parameters for LPC coefficients in wideband speech coding. In this paper, we propose a new hybrid coding scheme with multi-coder vector quantization for efficient coding of ISF parameters of the wideband speech coder AMR-WB. The coding system was designed based on four structured quantizers under noiseless channel conditions: the split vector quantizer (SVQ), the switched split vector quantizer (SSVQ), the multi-stage vector quantizer (MSVQ), and the multi switched split vector quantizer (MSSVQ). Simulation results show that our proposed AMR-WB ISF coding scheme outperforms conventional wideband ISF quantizers at lower bit rates.