节点文献
基于RBF网络和ADPCM的增益非线性量化
Gain Nonlinear Quantization Based on RBF Network and ADPCM
【Author】 Zhang Xueying Zhao Shuyan Zhang Gang(Taiyuan University of Technology, 030024, Taiyuan Shanxi, China)
【机构】 太原理工大学信息工程学院;
【摘要】 本文在G.728语音编码算法的码书归一化和增益精确表示的基础上,使用ADPCM对增益进行量化,提出对自适应量化阶距采用RBF网络进行非线性预测,将之运用到G.728算法的增益量化中。仿真结果表明:采用RBF网络的非线性自适应量化阶距方案比标准的G.728编码算法的平均分段信噪比提高1.63dB,这对于降低G.728算法的码率具有重要意义。
【Abstract】 This paper is based on the normalized shape codebook and gain described exactly of G728 speech coding algorithm, presents the principle of quantizating gain by ADPCM. The quantization step is predicted nonlinearly based on RBF network. They are used in gain quantizafion of G.728. The experimental results show that when new methods are used the average segment signal noise ratio (SNR) is increased by 1.63dB than that of G.728. It is significant to decrease code rate of G.728 while holding its other merits.
- 【会议录名称】 2004全国测控、计量与仪器仪表学术年会论文集(上册)
- 【会议名称】2004全国测控、计量与仪器仪表学术年会
- 【会议时间】2004-09
- 【会议地点】中国安徽黄山
- 【分类号】TN912.3
- 【主办单位】中国仪器仪表学会