Study On Hand-eye Interactive Technology For Intelligent Conference System

The Human-Computer Interaction(HCI) has attracted more and more researchers intelligent conference system. As new ways of HCI, eye tracking and gesture attract widespread attention. This paper proposes a gaze point estimation method based on spherical cross ratio and a Windowed Dynamic Time Warping joined side information method(SIWDTW) for gesture recognition, and uses this two methods in conference system. The main word content is as follows:According to traditional gaze point estimation method based on virtual plane, gaze point estimation method based on spherical cross ratio is proposed which overcomes the four virtual zero error for building virtual plane. This paper derives the relationship between spherical cross ratio and line cross radio, and then adopts the regression analysis to get the relationship. Finally, we can get the cross radio of screen and estimate the gaze point. Experiments show that this method has a better accuracy. Compare with the traditional method, accuracy of our method increases by 20.8% and 40.3%.We propose SJWDTW for gesture recognition. The windowed Dynamic Time Warping(WDTW) using a simple threshold to judge end points which is easily affected by noises. In order to overcome this drawback, we use probability judgment method instead of threshold judgment method, and join side information. The proposed algorithm experiments on our gesture database and the result is: The recall and precision of WDTW are 73.8% and 91.9%, while the recall and precision of SJWDTW are 87.6% and 89.9%. The experimental result shows that SJWDTW algorithm can improve the robustness.Finally, eye tracking and gesture recognition technology are used in the intelligent conference system to control computer which is more convenient and natural. Combine with eye tracking and gesture recognition technology, it can filter out the influence of speaker’s habitual action in speech and improve the robustness of the control greatly.