engleski

Towards Natural Head Movement of Autonomous Speaker Agent

Autonomous Speaker Agent (ASA) is a graphically embodied animated agent capable of reading plain English text and rendering it in a form of speech, accompanied by appropriate, natural-looking facial gestures [1]. This paper is focused on improving ASA’ s head movement trajectories in order to achieve facial gestures that look as natural as possible. Based on the gathered data we proposed mathematical functions that, using two input parameters (maximum amplitude and duration of the gesture) generate natural-looking head motion trajectory. Proposed functions were implemented in our existing ASA platform and we compared them with our previous head movement models. Our results were shown to a larger number of people. The audience noticed that results showed improvement in head motion and didn't detect any patterns which would suggest that animation was done with predefined motion trajectories.

head motion; gestures; HUGE; ASA; speaker agent

nije evidentirano