hide
Free keywords:
-
Abstract:
The ability to recognise complex movements and actions is critical for the survival of many species. In a series of psychophysical and functional imaging studies, we have investigated the role of learning for the recognition of complex motion stimuli. We trained human observers to discriminate between very similar human movement stimuli as well as between artificial movements. Our psychophysical results indicate no difference in the learning process for the two stimulus groups. Additionally, follow-up event-related fMRI adaptation experiments show an emerging sensitivity for the differences between the discriminated stimuli in lower-level motion-related areas (hMT+/V5 and KO/V3B). This effect was consistent for both stimulus groups. However, differences in brain activity between natural and artificial movements were obtained in higher-level areas (STSp and FFA). While sensitivity for artificial stimuli emerged only after training in these areas, sensitivity for natural movements was already present before training, and was enhanced after training. By extending a hierarchical physiologically inspired neural model for biological motion recognition (Giese and Poggio, 2003 Nature Reviews Neuroscience 4 179 - 192), we tried to model the BOLD signal changes during discrimination learning. The learning of novel templates for complex movement patterns was implemented by a combination of time-dependent and competitive hebbian learning, exploiting mechanisms that are physiologically plausible. The model accounts for the emerging sensitivity for novel movement patterns observed in fMRI. Learning of biological-motion patterns might thus be explained by a combination of several well-known neural principles in visual cortex.
