论文部分内容阅读
文中为解决超长帧编码系统中联合帧数过多导致LSP参数量化性能下降这个问题,将多项式拟合引入超长帧编码系统中,利用多项式拟合压缩LSP参数矩阵,以降低量化时实际联合的帧数。再用分裂矩阵量化器进行量化。仿真结果表明,在120 bps的参数量化速率下,此算法的量化性能优于矩阵量化算法(MQ),平均谱失真比MQ算法降低了0.0719 dB,平均分段信噪比提高了0.1971 dB。
In order to solve the problem of LSP parameter quantization performance degradation caused by the excessive number of joint frames in the super-long frame coding system, the polynomial fitting is introduced into the super-long frame coding system and polynomials are used to fit the LSP parameter matrix to reduce the actual joint The number of frames. Then use the split matrix quantizer to quantify. The simulation results show that the quantization performance of this algorithm is better than that of the matrix quantization algorithm (MQ) under the parameter quantization rate of 120 bps. The average spectral distortion is reduced by 0.0719 dB and the average segment SNR is improved by 0.1971 dB compared with the MQ algorithm.