查看文章

Action perception, planning and execution is a broad area of study, crucial for future development of clinical therapies treating social cognitive disorders, as well as for building human-computer interaction systems and for giving foundation to an emerging field of developmental robotics. We took interest in basic mechanisms of action perception, and as a model area chose dynamic perception of body motion. The focus of this thesis has been on understanding how perception of actions can be manipulated, how to distill this understanding experimentally, and how to summarize via numerical simulation the neural mechanisms helping explain observed dynamic phenomena. Experimentally we have, first, shown how a careful manipulation of a static object depth cue can in principle modulate perception of actions. We chose the luminance gradient as a model cue, and linked action perception to a perceptual prior previously studied in object recognition – the lighting from above-prior. Second, we have explored the dynamic relationship between representations of actions that are naturally observed in spatiotemporal proximity. We have shown an adaptation aftereffect that may speak of brain mechanisms encoding social interactions. To qualitatively capture neural mechanisms behind ours and previous findings, we have additionally appealed to the perceptual bistability phenomenon. Bistable perception refers to the ability to spontaneously switch between two perceptual alternatives arising from an observation of a single stimulus. Addition of depth cues to biological motion stimulus resolves depth-ambiguity. To account for neural dynamics as well as …