Journal of Computers Vol. 31 No. 2, 2020, pp. 12-19 doi:10.3966/199115992020043102002 12 Facial Expression Recognition Based on Deep Residual Network Junsuo Qu 1* , Ruijun Zhang 2 , Zhiwei Zhang 2 , Jeng-Shyang Pan 3 1 Xi’an Key Laboratory of Advanced Control and Intelligent Process, School of Automation, Xi’an University of Posts and Telecommunications, Xi’an 710061, China qujunsuo@xupt.edu.cn 2 Innovation Lab, Xi’an University of Posts and Telecommunications, Xi’an 710061, China {ruijunzhang, zzwft}@stu.xupt.edu.cn 3 School of Information Science and Engineering, Fujian University of Technolgy, Fuzhou 350000, China jengshyangpan@fjut.edu.cn Received 13 January 2019; Revised 7 March 2019; Accepted 12 March 2019 Abstract. Low accuracy of facial expression recognition for traditional methods, a facial expression recognition algorithm is proposed. Using the deep residual network model as the feature extractor, the residual block of the residual network is improved to enhance the information flow in the deep network. During training, apply some pre-processing techniques to extract only expression specific features from a face image and explore the presentation order of the samples and use softmax to classify and identify the extracted feature vectors. The experimental results show that a higher recognition rate is obtained on FER-2013. Keywords: deep residual network, facial expression recognition, pre-processing techniques, softmax 1 Introduction Facial expressions are the most natural way to reveal the inner world and play a vital role in our social interactions. Through facial expressions, you can express your feelings and infer others’ attitudes and intentions. Facial expression recognition (FER) is an essential part of the dynamic analysis and can be used to identify inner human emotions. FER methods attempt to classify facial expression in a given image or sequence of images as one of six basic emotions (anger, disgust, fear, happiness, sadness, and surprise) or as “neutral” [1]. Due to the complexity and subtlety of facial expressions and their relationship to emotions, accurate recognition of facial expressions still faces great difficulties. A typical FER system consists of three stages: (1) Face detection and localization. (2) Extract expression information from the located face. (3) A classifier (like an SVM) is trained on the extracted information to output the final expression labels. The facial expression recognition process is shown in Fig. 1. Fig. 1. Facial expression recognition process Traditional expression recognition is divided into three steps: feature learning, feature selection, and * Corresponding Author