Optimizing Feature Extraction using PSO-LDA for Face Recognition

Feature extraction is one of important process in face recognition LDA is dimensional reduction techniques that commonly used as feature extraction. Feature extraction by using LDA will produce feature space to extract important information of data. Selecting number of eigenvector which are used as feature space will not only effect on the computational time, but also effect on the recognition rate. This paper presents analysis number of eigenvector which is potential used as parameter extraction in feature extraction. The main idea of applying PSO in LDA is to search the number of parameter extraction for the optimal feature subset where features are carefully selected according to a well-defined discrimination criterion. Hybridizing PSO and LDA conducted to find the best number of feature space in order to optimize the recognition rate. The experiment conducted by using four databases with different disturbance, i.e. luminance, expression, focus and background, and also random disturbance. The results show that PSO-LDA can obtain the minimum number of parameter extraction which produces the highest recognition rate