B05 – Attention selection and recognition in scene-viewing

Scene viewing is an observable process of rapid decision making at the interface between perception, attention, and motor control. Knowledge about the organisation of the visual system has stimulated the development of attentional (“saliency”) models. Saliency models make predictions on regions that viewers are likely to fixate during scene exploration. Cognitive models for the control of eye movements study the dynamic interaction between perception and visuomotor control. Prior work predicts the distribution of fixation positions by a model of spatially-limited access to visual-saliency information and a leaky-memory model of re-inspection of positions. Besides macroscopic saccades, the eye performs involuntary micro-movements during saccades which are commonly categorized into micro-saccades, drift, and tremor.  It is known that macroscopic eye movements are highly individual, and eye movements have been explored as a mode of biometric identification. However, the accuracy of known approaches to oculomotor biometric identification is too low and the time that is needed for an identification is too long by one to two orders of magnitude for any practical application.

Project B05 has explored the roles of data and prior knowledge in the form of generative models of eye gaze for discriminative tasks such as viewer identification, and estimation of viewer familiarity. The first principal goal of the project was to develop dynamical, hierarchical Bayesian models of attentional selection in scene viewing that account for the viewers' familiarity with and conceptually-driven anticipation of the image content as well as for viewer-specific distributional properties of fixational patterns. The second principal goal was to exploit these generative models of fixation sequences as background knowledge within discriminative models that accurately predict values of latent variables, such as levels of familiarity, based on a given fixation sequence.