基于GERBF神经网络非线性预测的ADPCM算法控制系统中的应用  

A Nonlinear Prediction ADPCM Algorithm Based on GERBF Neural Network

作　　者：李志宏[1] 齐向东[1] 王安红[1] 

机构地区：[1]太原重型机械学院电子信息分院,太原030024

出　　处：《太原重型机械学院学报》2004年第3期181-184,共4页Journal of Taiyuan Heavy Machinery Institute

摘　　要：用神经网络建立非线性预测模型对语音信号进行处理,采用后向预测建模,不增加传输码率。采用一种改进的广义径向基函数网络(GERBF),利用正交最小二乘法训练速度快的优点,降低算法的复杂性。实验表明:基于GERBF预测器的语音编码系统在嵌入维数很少时亦能较好地去除语音信号相关性,其恢复语音质量优于CCITT,建议G.721中的ADPCM算法。A nonlinear prediction speech coding ADPCM algorithm based on RBF is introduced in this paper. In this coding system, the bit rate is not increased by the backward predictor. An improved General Radical Basis Function neural network (GERBF), which is trained by rapid Origin Least Square (OLS), is proposed. Compared with ADPCM algorithm based on BP, the computation is less evidently. Besides, the speech quality of the new system has an improvement of 1-2 dB than that of the speech coding standard G.721 system.

关 键 词：ADPCM 语音信号 码率 语音编码 语音质量 非线性预测 算法 神经网络 训练速度 径向基函数网络 

分 类 号：TH17[机械工程—机械制造及自动化] TN912[电子电信—通信与信息系统]