基于模糊支持向量机的语音识别方法

通过计算输入样本的模糊隶属度，探讨了模糊支持向量机（FSVM）的原理，应用其对语音信号进行识别。并和RBF神经网络、支持向量机（SVM）的识别效果进行了比较。在仿真实验中，采用小波分析方法提取语音特征向量，识别结果表明，SVM和FSVM比RBF网络具有较好的泛化性能，训练时间也大大缩减。此外，FSVM比SVM有更强的抵抗噪声的能力。

Speech Recognition Based on Fuzzy Support Vector Machine

The fuzzy membership to each input sample is calculated, and the principle of fuzzy support vector machine(FSVM) is discussed. Then FSVM is applied to recognize speech, and its classifying ability is compared with that of RBF network and support vector machine (SVM). During simulation experiment, wavelet analysis technique is adopted to extract feature vectors of speech, the results show that SVM and FSVM have both higher correct recognition rate and shorter training time than RBF network. Furthermore, FSVM is proved to have stronger ability to resist noise in input training samples.