High efficient quantization of the energy parameter in 0.6 kb/s vocoders
Tang Kun · Journal of Tsinghua University(Science and Technology) · 2008
Subjective tests show that the energy quantization precision influences synthesized speech quality in the 0.6 kb/s vocoder. An algorithm based on predictive vector quantization and dynamic allocation of bit resources improves the energy quantization precision by utilizing the fact that the pitch period is constant in unvoiced frames so that some of the bits originally allocated for pitch quantization can be re-allocated for energy quantization. The dynamically allocated bits are used to construct a second-stage codebook for better quantization of the energy parameter. Test results show that the algorithm reduces the quantization error by 70% compared with the gain-shape-based quantization algorithm.