【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。更多还原

【Abstract】 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.更多还原