Exploiting LCSVC Algorithm for Expression Recognition
Shuren Zhou, Ximing Liang, Can Zhu · 2008
Facial expression recognition basically requires fast processing speed as well as quality classification results. In this paper, an approach is presented for such a facial expression recognition using Locally Constrained Support Vector Clustering (LCSVC) and neural network (NN). During feature extracting, the independent component analysis can not only reduce the dimension of expression data, but also improve the clustering procedure of training data. Describing the LCSVC method in terms of Mixture of Factor Analysis (MFA) adjust the parameters of decision function mutually, it controls the some disturber of the clustering boundary, and also explains this method is better interpretable clusters than using support vector clustering (SVC) alone. The clustering result is further to construct a LCSVCNN. Using the support vectors, the LCSVCNN can determine where an unknown data belongs to one cluster that has been already built in the network. Experimental results prove the effectiveness of the proposed method.