提出了一种支持矢量机的汉语声调识别新方法。论文首先在基频和对数能量的基础上，建立了一个适合于支

持矢量机分类的等维声调特征。然后对支持矢量机的多分类策略和不同核函数对声调识别的影响进行了实验研究。

与BP神经网络相比，支持矢量机具有更高的识别率和更强的推广能力。-This paper presents a novel support vector machine based Chinese tone recognition method．A

new tone recognition feature is first ex血acted using the fundamental frequency(FO)and logarithmic

energy．And how to select the method of SVM multi-class classification and kernel function is also discussed

by experiments．Compared with BP neural network，SVM has higher recognition rates and more strong

generalization．