Fractional Supervised Orthogonal Local Linear Projection

In this paper, a subspace analysis method called the orthogonal local linear projection (OLLP) is proposed. OLLP is an unsupervised linear dimensionality reduction method with orthogonal basis functions. OLLP aims to find the projective map that optimally preserves the local structure of the data set. It shares many of the data representation properties of nonlinear techniques and resolves the out-of-sample problem. Furthermore, a fractional supervised variation on OLLP is also proposed by utilizing the class label information. Experimental results show that the proposed methods are effective for linear dimensionality reduction and achieve high recognition accuracy in facial expression recognition.