基于深度学习的面部微表情识别

Facial Microexpression Recognition Based on Deep Learning

微表情图片具有高度的相似性和密集性的细节信息,传统的微表情识别主要采用手工制作的方法,其识别种类与识别精度均无法满足精确的特征提取训练,因此提出一种深度学习方法,可以有效解决微表情识别在种类数量、准确度和速度上的问题.采用数据增强后合成的图像训练数据集,以处理后的数据集来训练卷积神经网络CNN模型.实验结果证明了所提出的基于深度学习的CNN方法在面部微表情识别中的有效性.将该方法与传统方法进行比较,结果显示提出的基于深度学习的CNN方法相较于传统方法其识别精度明显提高.

Micro-expression pictures have high similarity and intensive detail information.Traditional micro-expression recognition mainly adopts manual method,and its recognition type and recognition accuracy can not meet the accurate feature extraction training.Therefore,a deep learning method is proposed,which can effectively solve the problems of the type quantity,accuracy and speed of micro expression recognition.The image training data set synthesized after data enhancement is used,and the convolution neural network CNN model is trained with the processed data set.Experimental results show the effectiveness of the proposed CNN method based on deep learning in facial micro expression recognition.Comparing this method with the traditional method,the results show that the recognition accuracy of the proposed opportunity deep learning CNN method is significantly improved comparing with the traditional method.