Research Of The Facial Expression Recognition Based On The Embedded System

The research of facial expression recognition is combined with several disciplines, which can take a good role in every field. The application of expression recognition is every broad, which can be used in distance education, safe traffic, the medical profession and so on, so the research of expression recognition is of great significance.The thesis has researched various stage of facial expression recognition. By detecting and locating the original image information of people’s face, and then extracting and processing dimensionality reduction of the area of the detected face, at last, the scope field of recognized expression can be effected classified. The thesis highlighted introduce the facial test which based on the completion, the feature extraction which based on the two-dimensional Gabor wavelet, the feature of dimensionality reduction which based on the fusion algorithm of PCA and LDA, and the classification of expression recognition which based on SVM.In the feature dimensionality reduction, a new method was put forward to reduce the dimension of expression by combining the method of within the classes of PCA and the classical LDA. The result of the experiment shows that the method improved the speed of the computer to execute the algorithm and the recognition greatly.Based on the algorithm of people’s face, in order to meet the needs of different occasions, the thesis do some research on the basic of embedded platform Based on this platform which have features of low-power, a single process, low-cost, easy-take, easy-operate and so on, it also based on the core processor of ARM Cortex-A8. The platform main contains information of face images’collection driver and controller module, images’testing module and network transmission module. The thesis highlighted introduces the application of ARM Cortex-A8, image sensor ov5640, image acquisition interface video4linux and network transmission WG7310.The results of experiment shows that the designed platform has the features of stable performance, capturing the clear and accurate image data in time, and the practicality on the remote image information collecting.In experiment, the thesis collected the image information in different place on the embedded platform, combined the new algorithm, and did some testing and simulation on Visual C++and Mat lab.The result of simulation shows that the collected high clear image information can be tested the area of face effectively; the simulation of the fusion of algorithm of dimensionality reduction can improve the recognition of expression greatly.