| Facial expression recognition is based on Digital Image Processing, PatternRecognition and Computer Vision, it has important applications in human-computerinteraction and machine intelligence. But now only a few successful commercialinstances of facial expression recognition, this is mainly because it is still in theprocess of development and the related theories still need to be further studied. Butthe facial expression recognition will have a wide application prospect in the medical,security, public service and so on. It will lead to a new revolution in people’s lives.This paper uses Compressed Sensing (CS) and Support Vector Machine (SVM)for facial expression recognition. CS builds upon the fundamental fact that we canrepresent many signals using only a few non-zero coefficients in a suitable basis ordictionary. Then use some algorithms to classify or recovery of such signals from veryfew measurements. The SVM is based on the Statistical Learning Theory, and it has agood performance for small sample, nonlinear and high-dimensional classificationproblems. On the basis of the two theories, this paper carried out the following works:1. In this paper, the experiments selected880image samples from JAFFE,CAS-PEAL and Cohn-Kanada facial expression database. After the geometricpretreatment and gray pretreatment, those samples will be used for facial expressionfeatures extraction.2. Using the CS to extract facial expression features. The premise of the CS is thesparse of the signal, with these features CS can use redundant dictionary to sparserepresent the signal. Then use the sparse measurement matrix to observe the sparsesingle, after this process, the experiment can get the observation vector. Then use theobservation vector to constitute the feature vector. After the normalization, the featurevector can be used as classification characteristics of SVM.3. Using the SVM for facial expression recognition. Through differentexperiments explore the influence of the number of the features, kernel function, parameter optimization, classification categories to the classification accuracy rate, soas to find the best condition of facial expression recognition. Finally, compare thispaper method with the facial expression recognition based on Gabor and SVM toverify the superiority of this method on the classification accuracy and timeconsuming. |
