Boosted Independent Features For Face Expression Recognition

Lh He,Jz Zhou,D Hu,Cr Zou,L Zhao
DOI: https://doi.org/10.1007/11427445_23
2005-01-01
Abstract:Independent Component Analysis (ICA) is used widely to extract statistical independent features for analysis and discrimination in recent years. But its random properties make it very difficult to test the efficiency and. validation of the extracted independent features. In this paper, we propose a new method called BoostedICA to solve such problems by running ICA several times and boosting the selected independent components. Because of the local extremum question in calculating independent component, several times of running could get the more valid components with larger probability. The AdaBoost algorithm can guarantee the discriminating efficient of the selected features from the statistical theory. The proposed method achieves both computational efficiency and accuracy through optimizing extracting and choosing features. Finally we describe face expression recognition experiments on person-dependent and person-independent. The experimental results of 97.5% and 86% recognition rate respectively show that our method has better performance compared with other methods.
What problem does this paper attempt to address?