【摘要】 针对用于支持向量机的低维输入数据空间向高维特征空间的映射,通过黎曼测度张量扩大了支持向量机的线性可分边界,进一步提高了支持向量机分类的准确性。考虑到MLS-SVM的多分辨逼近效果和改进高斯核函数对支持向量机分类准确度的提升,企图努力给出一种基于两者优点的人脸表情识别算法,以反映人类在自然界中的认知过程,提出了采用改进高斯核的MLS-SVM人脸表情识别算法。实验结果表明,其人脸表情识别性能通过修改高斯核函数获得了较大的提升。更多还原

【Abstract】 Aiming at the map of the low dimension data space for support vector machines(SVM)to high dimension feature space,the Riemann measure tensor was used to extend the SVM linear disjunctive borders,and the SVM classification accuracy was improved.In fact,the good multiple resolution approximation accuracy of MLS-LSM has the similar process to that of the face emotion identification of human eyes.Therefore,one face emotion identification algorithm with the improved MLS-SVM by Guass kernel function was proposed in this paper,which could improve the face emotion identification accurate.At last,the numerical evaluation results show that the accurate of face emotion identification is improved with the comparison to two classification approaches.更多还原