Identifying Fixation and Saccades in Virtual Reality

Gaze recognition can significantly reduce the amount of eye movement data for a better understanding of cognitive and visual processing. Gaze recognition is an essential precondition for eye-based interaction applications in virtual reality. However, the three-dimensional characteristics of virtual reality environments also pose new challenges to existing recognition algorithms. Based on seven evaluation metrics and the Overall score (the mean of the seven normalized metric values), we obtain optimal parameters of three existing recognition algorithms (Velocity-Threshold Identification, Dispersion-Threshold Identification, and Velocity & Dispersion-Threshold Identification) and our modified Velocity & Dispersion-Threshold Identification algorithm. We compare the performance of these four algorithms with optimal parameters. The results show that our modified Velocity & Dispersion-Threshold Identification performs the best. The impact of interface complexity on classification results is also preliminarily explored. The results show that the algorithms are not sensitive to interface complexity.