Dynamic Facial Expression Recognition Based on Trained Convolutional Neural Networks
Recently, dynamic facial expression recognition in videos receives more and more attention. In this paper, we propose a method based on trained convolutional neural networks for dynamic facial expression recognition. In our system, we improve Deep Dense Face Detector (DDFD) developed by Yahoo to reduce training parameters. The LBP feature maps of facial expression images are selected as the inputs of the designed network architecture which is fine-tuned on FER2013 dataset. The trained network model is considered as a feature extractor to extract the features of inputs. In an image sequence, the mean, variance, maximun and minimum of feature vectors over all frames are calculated according to its dimensions and combined into a vector as the feature. Finally, Support Vector Machine is used for classification. Our method achieves a recognition accuracy of 53.27% on the AFEW 6.0 validation set, surpassing the baseline of 38.81% with a significant gain of 14.46%. The experimental results verify the effectiveness of our method.
Dynamic facial expression recognition Face detection Convolutional neural networks Local Binary Patterns Support Vector Machine
Ming Li Zengfu Wang
Institute of Intelligent Machines,Chinese Academy of Sciences,Hefei,Anhui,China;University of Science and Technology of China,Hefei,Anhui,China;National Engineering Laboratory for Speech and Language Information Processing,Hefei,China
国际会议
广州
英文
218-226
2018-11-23（万方平台首次上网日期，不代表论文的发表时间）