Further experiments in classification of attentional sequences: combining instantaneous and temporal evidence

Physiological studies on human visual system has shown that selective processing of visual information is one of the key features that lead to robust real time vision performance in a complex dynamic environment. Active vision systems simply simulate the selective attention behaviour of the eye, and work on the resulting spatio-temporal data to extract useful information in real time. In this work, we extend a previously proposed method for modelling and recognition of attentional sequences, based on the theory of evidence, and evaluate its performance.