论文部分内容阅读
８ｋｂｉｔ／ｓ短延时语音编码算法ＬＤ－ＡＣＥＬＰ，采用了代数码本激励线性预测（ＡＣＥＬＰ）的编码方法，利用语音的帧间相关性对线谱对参数采用了分裂式矢量量化技术，并采用高效的码本结构、码本搜索技术和增益矢量量化技术来获得较高的语音合成质量和较短的算法延时。ＬＤ－ＡＣＥＬＰ的帧长为１０ｍｓ，算法延时为１５ｍｓ。通过信噪比及人耳主观听觉实验等性能测试表明，该算法具有与国际电联１６ｋｂ／ｓ短延时语音编码算法ＬＤ－ＣＥＬＰ（Ｇ．７２８）相当的语音合成质量。
The 8 kbit / s short-delay speech coding algorithm LD-ACELP adopts the encoding method of Algebraic Code Excited Linear Prediction (ACELP), uses the inter-frame correlation of speech to split spectrum vector quantization and adopts the split vector quantization Efficient codebook structure, codebook search technology and gain vector quantization technology to obtain higher speech synthesis quality and shorter algorithm delay. LD-ACELP frame length of 10ms, the algorithm delay of 15ms. Performance tests such as signal-noise ratio and human ear subjective auditory experiments show that this algorithm has a speech synthesis quality equivalent to that of the ITU-T 16 kbps speech coding algorithm LD-CELP (G.728).