论文部分内容阅读
为了适应无线通信等低速语音编码场合 ,吸收 2 .4kb/s MELP算法优点 ,提出了一种改善激励源的 1 .2 kb/s混合激励线性预测算法 ,该算法在模型结构、LSP量化及解码等方面较传统的 2 .4kb/s LPC算法有了较大改进。同时 ,选用 TMS3 2 0 VC5 4 1 6 DSP芯片实时实现了以该算法为核心的声码器。客观音质测试和非正式主观测试结果表明 ,该算法在自然度、清晰度和抗噪声等方面明显优于 2 .4kb/s LPC算法 ,是低速率下一种良好的编码方案。
In order to adapt to low speed speech coding such as wireless communication and to absorb the merits of the 2.4kb / s MELP algorithm, a new proposed 1. 2 kb / s hybrid excitation linear prediction algorithm is proposed. This algorithm is based on model structure, LSP quantization and decoding And so on the more traditional 2.4kb / s LPC algorithm has been greatly improved. At the same time, we choose TMS320 VC5 4 1 6 DSP chip to realize the vocoder with the algorithm as the core in real time. Objective sound quality test and informal subjective test results show that this algorithm is obviously superior to 2.4kb / s LPC algorithm in terms of naturalness, sharpness and anti-noise, which is a good coding scheme at low speed.