Evaluation of a Case-based Facial Action Units Recognition Approach

Abstract

In this paper, we evaluate the performance of a case-based automatic facial action units recognition approach using interactive genetic algorithm (IGA). First, the case-based facial action units recognition approach is introduced. This method retrieves the most similar case image from case database using IGA and reuses the action units of the matched case image to the test face image. Second, to evaluate the effectiveness of our approach, comparison experiments with eigenface method on simple test images are done. The experimental results show that, for our method, the average recognition rate is about 77.5% on single AUs and average similarity rate is 82.8% on AU combinations, which are both higher than those of the eigenface method. Third, experiments of the case-based automatic facial action units recognition approach on complex test images is presented in this paper. The results prove the robusticity of our approach. A recognition rate of single AUs of 82.8% and a similarity rate of AU combinations of 93.1% are obtained