Research On User Emotion Recognition Based On Mouse Track And Face

Computer technology is developing rapidly,people can’t live without computers everywhere,and people’s requirements for computer interaction ability are gradually increasing,Therefore,the analysis and recognition of emotion has become an important interdisciplinary research topic in neuroscience,psychology,cognitive science,computer science and artificial intelligence.Through research,it is found that the design of website or game interface has a significant impact on user experience.The quality of interface design directly affects the mood of website or game users,and even affects the amount of users.Therefore,the emotion recognition of mouse users is carried out by mouse trajectory combined with face can better promote the improvement of website or game interface,is a relatively new research direction,and has certain research significance and commercial value.After understanding the current situation of mouse users’ emotion recognition at home and abroad,firstly,collect the unique data set of this paper.Secondly,three network models are used to train mouse trajectories through transfer learning to recognize and analyze mouse users’ emotions.Thirdly,propose a new network model for emotion recognition based on mouse trajectories,and compares the effect with the transfer learning network model.Finally,a network model is built for face images to realize emotion recognition and then prove the feasibility and credibility of emotion recognition through mouse trajectory.The main research contents are as follows:First,data collection.Crawl the positive and negative image sets in GAPED,design the experiment of emotion induction,collect the mouse track image and the face image of the experimenter under the same emotional tag at the same time,to establish a unique data set in this paper.Second,through transfer learning,VGG-16,Res Net50 and Mobile Net-V3-Large are used to train and learn the mouse track image collected in this paper to realize emotion recognition.The three network models have achieved good recognition results,the accuracy of VGG-16 recognition is 0.95,Res Net-50 recognition is 0.93,and Mobile Net-V3-Large recognition is 0.97.The correlation between mouse track and mood is verified,and the feasibility of recognizing mood by mouse track is verified.Third,Build a convolution neural network model to train and learn the mouse track dataset collected in this paper to achieve emotional recognition.The recognition effect is about 0.95,Compared with the three network models of migration learning,the recognition effect is similar,but the time-consuming is much shorter.Therefore,the four-layer convolution network model built in this section is more suitable for emotional recognition by mouse tracks,and it also proves that emotional recognition based on mouse tracks is feasible.Fourth,A convolution neural network model is proposed to train and learn collected at the same time as mouse track data and belonging to the same emotional tag to achieve emotional recognition.The training accuracy reaches about 0.9.Verify the correlation between face and emotion,and also indirectly verify the correlation between mouse track and emotion,which greatly increases the reliability of emotion recognition based on mouse track.