Summary: |
The Biomotion project main goal is to integrate knowledge from the fields of psychophysics, neurophysiology and computational theory on the multidisciplinary scientific theme of Vision. Under the common subject of the perception of biological motion, we propose to study the human visual perception system, collect psychophysical and neurophysiological data, investigate computational vision models and design biologically inspired integrated electronic circuits.
The rationale of this project is as follows: 1) first, we will use known and established paradigms of visual motion perception to devise and design several psychophysical experiments; 2) the perception experiments will be carried out in a virtual reality immersive environment with human subjects and their perceptive responses, such as reaction delays, will be recorded together with other neurophysiological variables such as evoked response potentials, thereby enabling further understanding of cortical chronometry of motion processing; 3) for some of the experiments, functional magnetic resonance imaging will be used in normal subjects to map human visual cortical circuits involved in biological motion perception; 4) the stimuli will be used as inputs to unsupervised learning algorithms, formulated according to high-order algebraic methods, capable of characterizing their multidimensional nature; 5) from the outputs of these methods, we expect to obtain sparse perceptive codes or neural coding "atoms" that match or, at least, correlate to specific human perceptive responses and, therefore, devise and propose new computational models that enclose the complexity of biological motion perception; 6) inspired in those models, dynamic image processing and analysis algorithms will be designed and the ones that best fit will be implemented in mixed signal integrated electronic circuits, in order to achieve computational and energetic efficiency; 7) finally, this artif |
Summary
The Biomotion project main goal is to integrate knowledge from the fields of psychophysics, neurophysiology and computational theory on the multidisciplinary scientific theme of Vision. Under the common subject of the perception of biological motion, we propose to study the human visual perception system, collect psychophysical and neurophysiological data, investigate computational vision models and design biologically inspired integrated electronic circuits.
The rationale of this project is as follows: 1) first, we will use known and established paradigms of visual motion perception to devise and design several psychophysical experiments; 2) the perception experiments will be carried out in a virtual reality immersive environment with human subjects and their perceptive responses, such as reaction delays, will be recorded together with other neurophysiological variables such as evoked response potentials, thereby enabling further understanding of cortical chronometry of motion processing; 3) for some of the experiments, functional magnetic resonance imaging will be used in normal subjects to map human visual cortical circuits involved in biological motion perception; 4) the stimuli will be used as inputs to unsupervised learning algorithms, formulated according to high-order algebraic methods, capable of characterizing their multidimensional nature; 5) from the outputs of these methods, we expect to obtain sparse perceptive codes or neural coding "atoms" that match or, at least, correlate to specific human perceptive responses and, therefore, devise and propose new computational models that enclose the complexity of biological motion perception; 6) inspired in those models, dynamic image processing and analysis algorithms will be designed and the ones that best fit will be implemented in mixed signal integrated electronic circuits, in order to achieve computational and energetic efficiency; 7) finally, this artificial vision system will be evaluated, according to its responses and performance with respect to the biological movement pattern recognition. If properly formulated and biologically plausible, the computational models proposed can further be used to design new psychophysical experiments to confirm hypothesis about mechanisms in biological motion perception and effectively closing the loop.
The psychophysical studies aim to understand the framework implied on the biological motion perception. The role of the dorsal and ventral visual cortex streams in biological motion detection and recognition, the nature of the representation level of the motion agent, the nature of the interaction observer - other agents - environment influence, cross-modal auditory visual integration and selective categorical attention phenomena will be studied. Starting with the traditional point light walker stimulus, we will manipulate structural complexity of the motion agent, maintaining the global translational components. Then, by using different methodologies (from psychophysics to immersive reality tasks) we will focus on the role of environment and motor schemes. The tasks of computational vision modelling and biologically inspired electronic design overlap in time with the psychophysical studies, because it is essential the research teams interact to achieve the goals of the project. |