Research On Visual Analysis Of Eye Movement Data For Danmaku Video

The appearance of danmaku enhances the interactive experience of the user,but also covers the video to a certain extent and affects the viewing experience.Therefore,understanding the impact of different danmaku styles on users’ viewing is important to generate a better danmaku design and enhance the viewing experience.Eye-tracking technology is a powerful tool for user experience evaluation by tracking users’ gaze points and eye movements,which can help analysts understand which elements affect users’ attention.However,the existing eye movement data analysis methods are mostly based on the overall analysis of data and lack the exploration of data details.Therefore,in this paper,we design an interactive visual analysis framework for eye movement data to help analysts gain more insights into eye-movement data.Firstly,the eye movement data is preprocessed to generate a data stream that is convenient for gaze point recognition.On this basis,a gaze point recognition algorithm based on density clustering algorithm is proposed,namely the I-DBSCAN algorithm.Secondly,in view of the problem that the danmaku text corresponding to the gaze point is difficult to obtain,two solutions are proposed based on the real-time position of the danmaku and optical character recognition technology.One is to calculate the real-time position of the danmaku by combining the danmaku text,danmaku appearance time and end time,and match the corresponding danmaku content according to the position of the fixation point.The second is to obtain the danmaku content and location under the current time slice through optical character recognition technology,and then match the corresponding danmaku according to the location of the fixation point.Thirdly,in view of the lack of support for detail exploration in the existing visualization methods,visual design is carried out.Firstly,the file upload view is designed.By providing files such as danmaku videos and eye movement data,data analysis can be automatically realized and corresponding visualization views can be generated to reduce the complexity of analysts’ work;then,custom interest area view is designed,and the analyst can divide the region of the background stimulus according to the task requirements,so as to improve the analyst’s initiative;finally,multiple views combining eye movement indicators and spatio-temporal visualization improve the multi-perspective of analyzing eye movement data.Finally,eye movement experiments are designed and case studies are conducted to explore the knowledge behind the eye movement data,verify the practicality of the visualization system,and propose the future improvement directions of the system.