An affective telepresence system using smartphone high level sensing and intelligent behavior generation

Anew approach in telepresence robots via automatic behavior generation is proposed in this paper. The robot behavior is generated using the Smartphone high level sensing, and personality based mood transition. The current telepresence systems require the simultaneous presence of both communication parties. Therefore, due to the time and context differences in distant communication, the opportunity of connectedness is reduced. This is especially important for more intimate telecommunications, that ongoing connectedness is more required. To solve these problems, we developed an automatic behavior generation system, that produces behaviors on behalf of the remote person. In order to be able to infer the state of the remote person more frequently and dynamically, Smartphone sensors are used for automatic high level sensing. Furthermore, to produce more believable affective expressions by the robot, the expressions correspond to the remote user's personality stereotype. Fuzzy Kohonen Clustering Network (FKCN), is utilized to linearly fuse the inferred mood states and regenerate them on the agent. To evaluate the results 120 samples of the user's Smartphone usage and the mood in different times of the day were logged. The state estimated by the model were compared against the users self report data. The results showed that there was no significant difference between the user's selfperceived feelings, and the feelings recognized by the model.