The effect of blurring on action recognition by human subjects

Several experiments on human subjects evidenced how they are endowed with a good ability to recognize unambiguously a motor action, whilst they are not able to recognize without uncertainty an action from a single image frame representing the same action. In fact, the movement is a motion characterized by a whole spatio-temporal pattern, with specific spatial distributions of kinetic energy, and is sufficient for extracting meaningful information about 3-D shapes. In this contribution we study how the interaction between two systems (the movement image and the cognitive system of the observer) changes when a feature of the image is manipulated. The image features were its dynamics, its shape and the spatial relationships occurring between its components. In order to investigate the effects of these features we showed to 80 students movies and slides depicting different actions, with four blurring levels (0–3), obtained through a mosaic technique that masks shape and spatial relations. The subject task was to give a verbal interpretation of the action depicted by the pattern itself. The goals of the experiment were to know if humans were able to recognize actions only getting information from dynamics, and at what shape definition level a human observer was able to distinguish two movements that showed the same dynamics but were associated to different actions. The findings evidenced that the blurring level had a strong influence if the input pattern was a static image. So the movement carries a lot of information about the dynamics and the action that is quite independent from the objects contained in the scene. Moreover, in correspondence to the same blurring level, actions and dynamics were not recognized with the same precision, leading to a recognition of dynamics without the corresponding recognition of the associated action. This supports the hypothesis asserting the existence, within subjects, of general movement schemes operating in the recognition process