In this talk. we address the problem of audiovisual speaker tracking. We model the tracking problem as a system of linear equations and follow an approach which involves the incorporation of stream weights into the conventional Kalman filtering paradigm.  Initially. fixed stream weights are used to define a strategy to methodologically incorporate information from both acoustic and visual sensory modalities.   That approach is then augmented and optimized to allow for the incorporation of dynamic stream weights into the conventional multimodal Kalman filter.  We present all aforementioned methods along with simulations and results of our programs. Our results show that the adopted techniques efficiently and effectively combine acoustic information with visual cues to perform speaker tracking and that the method of incorporating stream weights outperforms the tracking capabilities of the conventional Kalman Filter.

Place: Hybrid: Math, 402/Zoom:   Link  https://arizona.zoom.us/j/83541348598  Password:  BB2022