论文部分内容阅读
主观听觉测试表明,在0.6 kb/s声码器算法中,能量参数的量化精度与合成语音的质量有着密切的关系。为提高能量参数的量化精度,该文提出了一种基于预测矢量量化与比特资源动态分配的能量参数量化算法,在超帧编码过程中,充分利用基音周期在清音帧为固定值的特点,将其部分量化比特分配给能量参数,构造其二级码本,从而更好地量化能量参数。实验结果表明:与基于增益-形状的量化算法相比较,该文提出的算法可以使能量参数的量化误差降低70%左右。
Subjective auditory tests show that in the 0.6 kb / s vocoder algorithm, the quantization accuracy of energy parameters is closely related to the quality of synthesized speech. In order to improve the quantization accuracy of energy parameters, an energy quantization algorithm based on vector quantization and bit resource dynamic allocation is proposed in this paper. In the process of super-frame encoding, Part of its quantized bits are assigned to the energy parameters, and their secondary codebooks are constructed so as to better quantify the energy parameters. Experimental results show that compared with the gain-shape quantization algorithm, the proposed algorithm can reduce the quantization error of energy parameters by about 70%.