Facial Expression Recognition Based On Infrared Images

Recognition of infrared facial expression has important significance to people’sdaily life. The infrared facial image is independent of light, so it cannot be influenced bypeople’s skin colors and makeup. It can be used to do normal facial expressionrecognition in the night or sufficient of light, which has important significance to realizesmarter human-computer interaction.According to infrared facial expression recognition, we did the following work:(1) Infrared face feature location. Do organs of the face feature points detection usingHarris corner detection algorithm based on region segmentation.Position the key pointsof the eyes, nose, mouth, which provide many location information for further facialexpression recognition(2) Infrared facial feature extraction. On the basis of the feature location, use UniformLocal Binary Patterns (ULBP) algorithm to extract the texture features of the infraredface. This can extract effective texture information of infrared face.(3) Use PCA to do feature selection. Do feature dimension after the ULBP featureextraction, discard some features which contain little information. And we will reducethe number of dimensions of the characteristics. Reduce the facial feature dimensions onthe basis of keeping the recognition rate, and reduce computation as well as improveexpression recognition speed.(4) Infrared facial expression classification. Choose SVM classifier to do infrared facialexpression classification based on LBP feature extraction, which effectively prevent theclassifier may cause local minimization problem and linear inseparable. Finally ensuregood recognition accuracy.(5)Compared other feature extraction methods as well as other SVM classifiers usingother core function on the results of classification on infrared facial expressionrecognition. Finally, it is certified that the improved method is effective on this research.