Research And Integrated Realization Of Facial Expression Recognition

Facial expression recognition is an important research branch of biometric recognition, which can not only reflect the emotional and psychological change, but also has a natural and harmonious interaction. Research of facial expression recognition in this thesis has an important theoretic significance and widely application value in the aspects of human-computer interaction and emotional understanding.This thesis completes the overall design of facial expression recognition system, and methods of expression feature extraction and expression classification are the research emphasis. First of all, algorithm of AdaBoost is used to detect face accurately. And then, technology of image pretreatment is used to process facial image, which can reduce the influence of illumination variation and different size for expression feature extraction.In stage of expression feature extraction, it is considered that single method of expression feature extraction would ignore global or local expression features, this thesis proposes a fusion method of principal component analysis(PCA) and local directional pattern(LDP) for expression feature extraction. LDP is used to extract textural features of eyes and mouth areas where have much contribution to facial expression recognition. Then textural features are combined with global features extracted by PCA to obtain more effective expression features. The experimental results show that the method of fusing PCA with LDP can obtain higher expression recognition rate than single method.In stage of expression classification, method of DAGSVM(directed acyclic graph support vector machine) uses euclidean distance to calculate the degree of separation between different classes, and this would cause problem of error accumulation easily. Therefore, an improved method of DAGSVM based on relative distance is proposed to solve the problem of error accumulation. The experimental results show that the improved method of DAGSVM can solve the problem of error accumulation effectively, and improve the accuracy of expression classification.Finally, facial expression recognition is integrated to motion control system of the intelligent wheelchair in this thesis. In platform of intelligent wheelchair, hardware system is made up of ARM processor, machine controller, motor driver and so on. Then open source computer vision library(OpenCV) and microsoft foundation classes(MFC) are used for programming. And then, different expressions recognized in this thesis are transformed into different instruction for controlling the movement of intelligent wheelchair. The experimental results show that the method proposed in this thesis can recognize different expressions and control the movement of intelligent wheelchair effectively, and has good stability at the same time.