Analysis of Behaviors by Audience in Lectures by Using Time-series Models
Watanabe Eiji, Takashi Ozeki, Takeshi Kohama · 2014
In this paper, the dominant behaviors define by the face direction of the speaker and audience in lectures are analyzed by using the time-series models. First, we detect the number of skin-colored pixels in face region of speaker and audience as features for behaviors by them. Next, we construct piecewise time series models for these features for the speaker and audience. Finally, we analyze the synchronization phenomena in speaker and audience by comparing time series models. Moreover, we discuss the relationships among notes, test and behaviors by audience.