Micro-expression Recognition Based On Deep Convolutional Neural Network

Micro-expression is a kind of brief facial expression which could not be controlled by nervous system.Micro-expression indicates that a person is hiding his truly emotion consciously.Micro-expression recognition has various potential applications in public security and clinical medicine.Some researches are focused on the automatic micro-expression recognition,because it is hard to recognize the micro-expression by human eyes.The main works of this thesis are illustrated as follow:1.This thesis proposed a novel algorithm for automatic micro-expression recognition which combined a deep multi-task convolutional network for detecting the facial landmarks and a fused deep convolutional network for estimating the optical flow features of the micro-expression.Firstly,the deep multi-task convolutional network is employed to detect facial landmarks with the manifold related tasks for dividing the facial region.Furthermore,a fused convolutional network is applied for extracting the optical flow features from the facial regions which contain the muscle changes when the micro-expression presents.Because each video clip of micro-expression has many frames,the original optical flow features of the whole video clip will have high dimension and redundant information.This thesis revises optical flow features for reducing the redundant dimensions.Finally,a revised optical flow feature is applied for refining the information of the features and a Support Vector Machine classifier is adopted for recognizing the micro-expression.The result of experiments on two spontaneous micro-expression database demonstrate that the method proposed in this thesis achieved good performance in micro-expression recognition.2.Because there are some samples of a micro-expression database which contain human face with the small size,this thesis introduces the algorithm of Single Image Super-Resolution into the micro-expression recognition.This thesis proposes a deep convolutional neural network model with shuffle dense connections,named SDSR,for reconstructing the low resolution input image into the high resolution image.Specifically,dense skip connections is adopted for combining the low-level features and the high-level features to enhance the performance of the reconstruction and residual learning is applied for easing the difficulty of training the deep neural network.In addition,the grouped convolutions are introduced for reducing computational complexity and the number of parameters.What's more,the shuffle dense connections are proposed for mitigating the grouped convolutions problem of lacking information exchange between the groups.The proposed network is evaluated in publicly available databases and the results illustrate that this approach has great performance in SISR.This thesis utilizes this network for reconstructing these samples which contain human face with the small size into the high resolution images and utilizes these reconstructed images for micro-expression recognition.The result of experiments on the micro-expression database demonstrate that the proposed method with reconstruction processing achieved better recognition performance than original method.