Person-independent facial expression recognition based on person-similarity weighted distance

In this paper, a new distance called person-similarity weighted distance is proposed for person-independent facial expression recognition. In the new method, expression features associated with all persons in training set are extracted by Higher-Order Singular Value Decomposition (HOSVD) firstly. Then, based on the assumption 'similar persons have similar facial expression', the person-similarity weighted distance is calculated to measure the similarity between the test expression and standard expressions. By the weighting process, the distance can remove the differences caused by individual and becomes less person-dependent. Experimental results show the superiority of proposed method over the existing methods.