摘要
分析研究了增益、波形乘积码书结构的缺陷,设计了归一化波形码书和精确表示增益的LD-CELP方案。采用自适应预测和自适应量化对增益的精确值进行量化,在3 bit 和4 bit自适应量化时比G.728固定量化增益分别提高0.5dB和6 dB。采用4 bit自适应量化和64波形码书比G.728 SNR提高约1dB。将G.728综合滤波器由50阶减少到30阶,信噪比不变而算法复杂性降低约20%。
The limitation of gain and shape codebook structure was analyzed and studied, and a LD-CELP scheme was proposed by normalized shape codebook with exact gain, in which the exact gain value was adaptive predicted and adaptive quantized. In this method, for adaptive quantization gain of 3 bit and 4 bit, the SNR have been increased by 0.5 and 6 dB separately than that of G.728. If adopting 4 bit adaptive quantization and 64 codebook, the SNR had 1 dB improvement than G.728. When the order of synthesis filter is reduced from 50 to 30, the SNR isn't changed and the complexity of G.728 is decreased by 20%.
出处
《通信学报》
EI
CSCD
北大核心
2003年第9期87-92,共6页
Journal on Communications
基金
国家自然科学基金资助项目(60172070)
山西省自然科学基金资助项目(991040
20011033)