A Novel Isolated Speech Recognition Method Based on Neural Network
A Novel Isolated Speech Recognition Method Based on Neural Network
复制标题
一种基于神经网络的孤立语音识别新方法
DOI：
10.1007/978-1-4471-4844-9_58
复制
发表时间：
2013
期刊：
影响因子：
--
通讯作者：
Guojiang Fu
中科院分区：
文献类型：
--
作者：
Guojiang Fu
径向基函数神经网络结构已被证明适用于孤立词的识别。单词的识别以依赖于说话者的模式执行。在这种模式下，呈现给网络的测试数据与训练数据相同。每帧16个参数的16个线性预测倒谱系数，由于倒谱中的前16个系数代表了大部分共振峰信息，因此改进了一种很好的语音特征提取方法。结果表明，径向基函数神经网络(RBF)分类器的性能优于MLP分类器。结果表明，在训练MLP分类器时，6个说话人的平均表现最好，在训练RBF分类器时，说话人的2个平均表现最好。结果表明，在测试MLP分类器时，说话人4的平均性能最好，在测试RBF分类器时，说话人1的平均性能最好。
The Radial Basis Function Neural Network architecture has been shown to be suitable for the recognition of isolated words. Recognition of words is carried out in speaker-dependent mode. In this mode the tested data presented to the network are the same as the trained data. The 16 Linear Predictive Cepstral Coefficients with 16 parameters from each frame improves a good feature extraction method for the spoken words, since the first 16 in the cepstrum represent most of the formant information. It is found that the performance of radial basis function neural network (RBF) classifier is superior to MLP classifier. It is found that speaker 6 average performances is the best performance in training MLP classifier and speaker 2 average performances is the best performance in training RBF classifier. It is found that average speaker 4 performances is the best performance in testing MLP classifier and speaker 1 average performance is the best performance in testing RBF classifier.