Study On Some Problems Of Automatic Facial Expression Recognition

Facial expression recognition is one of the most challenging problems in the fields of pattern recognition, machine vision, affective computing and psychology. It has turned into an active research topic in the recent decades. Although facial expression automatic recognition technique develops quickly along with various applications, there are still many problems unsettled yet.Based on the former research achievements, some improved algorithms were proposed to overcome the difficulties of this topic. In this paper, we pay more attention to the following items.(1) Genetic algorithms applied to face detection were studied, which included multiple faces genetic detection algorithm, chaotic quantum genetic algorithm (CQGA) improved symmetry involved face detection, sparse features for face detection extraction with CQGA, and a multiple-scaled various classifications scheduled face detection algorithm. The simulation and experimental results show that the presented face detection algorithms perform well, and the GA improves the speed of features extraction or face detection. Especially, the multiple-scaled various classifications scheduled face detection algorithm takes information between frames, skin color segmentation, and eyes detection to get coarse results. And then, different features were used according to the different sizes of detection windows. Results show that this method outperforms the single strategy algorithms on both robustness and efficiency.(2) The three main Expression features extraction methods were studied including eigenflow, dense flow and wrinkle extraction.(3) Facial expression recognition based on chaos modulation and correlation detection was presented, which takes the chaos modulation rememberable character into use. Results show better recognition rate especially on peak facial expression than others involved. Hidden Markov Models (HMM) is also studied to recognize facial expressions with 93.5% correct rate.(4) The investigation aims at overcoming the shortcomings of current facial expression classification. Methods were presented to improve Facial Action Coding System (FACS) to form a fine one.