FACIAL EXPRESSION RECOGNITION BASED ON IMPROVED LBP OPERATOR AND K-MEANS CLUSTERING

For facial expression features extraction and automatic clustering, a method is proposed. This method integrates improved LBP operator and K-means clustering. Firstly, facial features are located by integral and variance projection. Improved LBP operator with changed threshold is figured out. Secondly, K-means clustering produces expression templates. Then K nearest neighbor algorithm recognizes facial expression. Meanwhile, the expression templates will be updated by the result itself, optimizing expression templates dynamically. The algorithm is simple and low computational-complexity. It updates the train results dynamically. Results show that this classification method is accurate.