LAUSR

Eye movement is proven to be the most frequent activities of human beings; as a result research on recognition of unit eye movement has become a hotspot in human activity recognition. In this paper, we propose a robust online saccade recognition algorithm, which integrates electrooculography (EOG) and video together. Initially, EOG signals and video data are collected simultaneously from eight saccadic directions. Then online active eye movement segment detection algorithm is developed to detect the effective saccadic signal from ongoing eyeball activities. Furthermore, we extract features from different modalities and explore two fusion strategies [i.e., feature level fusion (FLF) and decision level fusion (DLF)]. In laboratory environment, the average recognition accuracy of FLF and DLF achieves 89.37% and 89.96%, respectively, which reveals that the proposed method can improve the performance of consecutive saccade recognition in comparison with sole modality.