方向基函数神经网络在说话人识别中的应用  

Application of Direction-based Function Neural Networks in Speaker Verification

作　　者：徐春桥[1] 孙瑞晓[2] 史静朴[3] 

机构地区：[1]河北林业学校微机室,河北石家庄050061 [2]河北科技大学人事处,河北石家庄050018 [3]中科院半导体所神经网络实验室,北京100083

出　　处：《河北工业科技》2004年第1期26-28,36,共4页Hebei Journal of Industrial Science and Technology

摘　　要：首先对方向基函数神经网络模型(Direction basisFunctionNeuralNetworks)进行了描述。这种模型可以通过计算矢量间的夹角来对矢量进行相似性判别。与径向基函数神经网络模型(Radial basisFunctionNeuralNetworks)不同的是,它特别适合于具有方向不变性的模式识别领域,如语音识别。重点探讨了方向基函数神经网络在说话人识别中的应用,在此基础上研制成功了1个说话人确认系统并用硬件进行了实现。实验数据表明这个系统具有很好的性能。Direction-based function neural network model, which determines the similarity of two vectors through calculating their angles, is described. This neural network model, different from radial-based function neural networks, can be better used in such pattern recognition field as speech recognition, where the pattern of a vector is not determined by its amplitude,but by its angle. The application of direction-based function neural networks in speech recognition is further explored and a speaker verification system implemented by hardware is successfully developed. Experiments results indicate that this system has good performance.

关 键 词：神经网络 方向基函数 径向基函数 语音识别 说话人识别 说话人确认 

分 类 号：TN912.34[电子电信—通信与信息系统] TP183[电子电信—信息与通信工程]