作　　者： ; ;

机构地区： 华南理工大学电子与信息学院

出　　处： 《通信学报》 2000年第10期31-37,共7页

摘　　要： 为改善预测类声码器中长时预测器特性 ,本文引入了一种全连接回归神经网络 (FRNN)非线性预测器并将其应用于话音编码算法中。FRNN在隐层单元不仅有来自自身的反馈 ,也有来自输出单元的反馈 ,因此其预测性能好于常规预测器。将其应用于码本激励话音编码系统 (CELP)中 ,可得到较低的传输码率 ,同时亦可改善编码质量。 To improve the long term correlation prediction characteristics in speech coding,this paper propose a new nonlinear predictor,i e ,a fully connected recurrent neural network(FRNN) where the hidden units have feedbacks not only from themselves but also from the output unit The comparison of the capabilities of the FRNN with conventional predictors shows that the former has less prediction error than the latter We apply this FRNN in the code excited predictive speech coding system(CELP) and shows that the FRNN requires less bit rate and improves the performance for speech coding

关 键 词： 非线性预测 回归神经网络 话音编码 码本激励

领　　域： [电子电信] [电子电信]