Unsupervised Learning of Head Pose through Spike-Timing Dependent Plasticity
PIT '08 Proceedings of the 4th IEEE tutorial and research workshop on Perception and Interactive Technologies for Speech-Based Systems: Perception in Multimodal Dialogue Systems
Hi-index | 0.00 |
The goal of this work is to classify the focus of attentionof a subject who is switching his or her attention between anumber of surrounding objects. The specific application isto classify the focus of attention of a car driver as straight-ahead,towards the dashboard, or towards one of the mirrors.A quantitative approach to this problem requires (a) apriori information about the interior geometry of the carand the calibration of the camera, and (b) accurate computationof the subject's location and eye direction. Thispaper describes an alternative, qualitative, approach. Thedriver is observed over an extended period of time, andfrequently occurring view directions are identified. Eachof the frequent view directions is associated with its mostlikely target - dashboard, rear-view mirror etc. The headpose of the driver in any subsequent image is then classifiedby relating it to one of these labelled view directions.The head pose classification is augmented with a qualitativemeasurement of the eye pose.