LouposKostas
12-04-2007, 11:56 AM
Post-doctoral position
Project Title : Probabilistic approach to multi-sensorial integration in Virtual Environments.
Project managers : Bourdin Christophe, MCU Université de la Méditerranée and Mestre Daniel, DR CNRS
Research Project :
Context
Perceiving the surrounding world and the control of our actions (mainly here spatial orientation and navigation) is dependant upon the combination, at the Central Nervous System (CNS) level, of information derived from the different sensorial (visual, auditory, somesthetic) modalities. Hence, cortical and cognitive mechanisms contribute to the combination of various sensorial signals into a coherent and effective representation of the environment, the self and their interaction. These mechanisms are multi-sensory integration processes (Deneve and Pouget, 2004). Their nature, structure and functional characteristics are complex, for many reasons:
- First, the reliability of sensorial signals varies significantly as a function of the environmental context. For instance, in daylight, visual signals are more precise than auditory signals when subjects have to localize a stimulus, and the reserve occurs in darkness. Multi-sensorial integration research also has to take into account the fact that sensorial signals are, by essence, noisy, due to biological sensors properties (Körding and Wolpert, 2006 ; Roach et al., 2006). Sensorial noise and information reliability are closely related.
- Secondly, each sensorial modality involves its own neural coding mechanisms, in order to account for common environmental and corporal properties. The complementarity / redundancy of information delivered by different sensorial systems must be processed by the CNS, in order to build-up an unified perception of the body in the environment.
Until now, multi-sensorial integration mechanisms are poorly understood and modeled. Deciphering the mechanisms by which the human brain integrates various sensorial information into a coherent percept (a gestalt) is a fundamental question in the domain of Integrative Neuroscience. However, a number of "integration" models were previously proposed. The reliability problem (presented above) suggests the interest of a statistical approach, in which multi-sensorial integration is, fundamentally, a probabilistic inferential process. The hypothesis of a Bayesian encoding mechanism considers that the CNS represents sensorial information in the form of probabilistic distribution (van Beers et al., 1999 ; Knill, 2003). As a consequence, Bayesian formalisms might be efficient tools to model the integration of various sensorial signals, while accounting for their reliability levels.
Methodology
The main goal of this research project is to contribute to a better understanding of multi-sensorial (visual, auditory, vestibular and somesthetic) integration mechanisms in spatial orientation and navigation tasks. A Bayesian approach will be deployed, in order to describe transfer functions between sensorial inputs and a unified percept and behavior. To achieve this goal, virtual reality experimental systems are, by essence, multi-sensorial stimulation devices and are thus appropriate to study multi-sensorial integration. Virtual Reality devices enable experimental manipulation and control of ecologically valid stimulation. The recent development of increasingly sophisticated immersive systems, involving dynamic and stereoscopic visual stimulation, spatial sound, haptic stimulation (force-feedback systems), vestibular stimulation (centrifuge, galvanic stimulation, tactile stimulators) render possible the experimental investigation of the implication, respective role and interactions of the different sensorial systems in the perception and control of the self in the environment. All of the above mentioned experimental devices and systems are available and functional in different labs in our research institute.
Main objectives
On the basis of experimental data from human subjects, a functional model of the perception and control of self-orientation in space is sought for, in an attempt to deepen our understanding of complex multi-sensorial integration processes. A Bayesian approach will be used to address key (yet unsolved) issues in movement perception neuronal mechanisms. These models will also be applied to progress in the conception and tuning of automotive simulators and virtual reality systems. Indeed, simulator designers are facing a problem that is exactly reverse to the one we are addressing here: They have to induce a coherent movement percept by generating combinations of sensorial inputs. The candidate will have full access to experimental methods and devices of our institute (multidisciplinary expertise, "state-of-the-art" CAVE and HMD immersive systems, centrifuge, driving simulators).
Candidate profile
Such a multidisciplinary research project calls for a post-doc researcher who will be funded for 24 months. The candidate will have an expertise in the Bayesian statistical decision theory applied to perceptual systems. Competence in Cognitive Sciences AND Computational Sciences will be a clear advantage. The candidate will also have had concrete achievements in Experimental Psychology and Behavioral Sciences, and extensive knowledge of human sensorial systems. A background in virtual reality will also be considered as a plus. The candidate should have graduated in one of the following disciplines: Cognitive Sciences, Integrative Neuroscience, Computational Neuroscience, Ergonomics, Human Movement Sciences.
To sum up, required scientific expertise includes:
• Bayesian statistical decision theory
• Formal modeling of the Central Nervous System
• Experimental approach to human behavior
Such expertise will be demonstrated by a significant number of peer-reviewed scientific publications.
Contact :
Christophe Bourdin, MCU, Université de la Méditerranée, christophe.bourdin@univmed.fr, +33 (0)4 91 17 22 76
Daniel Mestre, DR CNRS, daniel.mestre@univmed.fr, +33 (0)4 91 17 04 38
UMR 6152 Mouvement et Perception
163 avenue de Luminy
CP 910
13288 Marseille Cédex 09
FRANCE
Project Title : Probabilistic approach to multi-sensorial integration in Virtual Environments.
Project managers : Bourdin Christophe, MCU Université de la Méditerranée and Mestre Daniel, DR CNRS
Research Project :
Context
Perceiving the surrounding world and the control of our actions (mainly here spatial orientation and navigation) is dependant upon the combination, at the Central Nervous System (CNS) level, of information derived from the different sensorial (visual, auditory, somesthetic) modalities. Hence, cortical and cognitive mechanisms contribute to the combination of various sensorial signals into a coherent and effective representation of the environment, the self and their interaction. These mechanisms are multi-sensory integration processes (Deneve and Pouget, 2004). Their nature, structure and functional characteristics are complex, for many reasons:
- First, the reliability of sensorial signals varies significantly as a function of the environmental context. For instance, in daylight, visual signals are more precise than auditory signals when subjects have to localize a stimulus, and the reserve occurs in darkness. Multi-sensorial integration research also has to take into account the fact that sensorial signals are, by essence, noisy, due to biological sensors properties (Körding and Wolpert, 2006 ; Roach et al., 2006). Sensorial noise and information reliability are closely related.
- Secondly, each sensorial modality involves its own neural coding mechanisms, in order to account for common environmental and corporal properties. The complementarity / redundancy of information delivered by different sensorial systems must be processed by the CNS, in order to build-up an unified perception of the body in the environment.
Until now, multi-sensorial integration mechanisms are poorly understood and modeled. Deciphering the mechanisms by which the human brain integrates various sensorial information into a coherent percept (a gestalt) is a fundamental question in the domain of Integrative Neuroscience. However, a number of "integration" models were previously proposed. The reliability problem (presented above) suggests the interest of a statistical approach, in which multi-sensorial integration is, fundamentally, a probabilistic inferential process. The hypothesis of a Bayesian encoding mechanism considers that the CNS represents sensorial information in the form of probabilistic distribution (van Beers et al., 1999 ; Knill, 2003). As a consequence, Bayesian formalisms might be efficient tools to model the integration of various sensorial signals, while accounting for their reliability levels.
Methodology
The main goal of this research project is to contribute to a better understanding of multi-sensorial (visual, auditory, vestibular and somesthetic) integration mechanisms in spatial orientation and navigation tasks. A Bayesian approach will be deployed, in order to describe transfer functions between sensorial inputs and a unified percept and behavior. To achieve this goal, virtual reality experimental systems are, by essence, multi-sensorial stimulation devices and are thus appropriate to study multi-sensorial integration. Virtual Reality devices enable experimental manipulation and control of ecologically valid stimulation. The recent development of increasingly sophisticated immersive systems, involving dynamic and stereoscopic visual stimulation, spatial sound, haptic stimulation (force-feedback systems), vestibular stimulation (centrifuge, galvanic stimulation, tactile stimulators) render possible the experimental investigation of the implication, respective role and interactions of the different sensorial systems in the perception and control of the self in the environment. All of the above mentioned experimental devices and systems are available and functional in different labs in our research institute.
Main objectives
On the basis of experimental data from human subjects, a functional model of the perception and control of self-orientation in space is sought for, in an attempt to deepen our understanding of complex multi-sensorial integration processes. A Bayesian approach will be used to address key (yet unsolved) issues in movement perception neuronal mechanisms. These models will also be applied to progress in the conception and tuning of automotive simulators and virtual reality systems. Indeed, simulator designers are facing a problem that is exactly reverse to the one we are addressing here: They have to induce a coherent movement percept by generating combinations of sensorial inputs. The candidate will have full access to experimental methods and devices of our institute (multidisciplinary expertise, "state-of-the-art" CAVE and HMD immersive systems, centrifuge, driving simulators).
Candidate profile
Such a multidisciplinary research project calls for a post-doc researcher who will be funded for 24 months. The candidate will have an expertise in the Bayesian statistical decision theory applied to perceptual systems. Competence in Cognitive Sciences AND Computational Sciences will be a clear advantage. The candidate will also have had concrete achievements in Experimental Psychology and Behavioral Sciences, and extensive knowledge of human sensorial systems. A background in virtual reality will also be considered as a plus. The candidate should have graduated in one of the following disciplines: Cognitive Sciences, Integrative Neuroscience, Computational Neuroscience, Ergonomics, Human Movement Sciences.
To sum up, required scientific expertise includes:
• Bayesian statistical decision theory
• Formal modeling of the Central Nervous System
• Experimental approach to human behavior
Such expertise will be demonstrated by a significant number of peer-reviewed scientific publications.
Contact :
Christophe Bourdin, MCU, Université de la Méditerranée, christophe.bourdin@univmed.fr, +33 (0)4 91 17 22 76
Daniel Mestre, DR CNRS, daniel.mestre@univmed.fr, +33 (0)4 91 17 04 38
UMR 6152 Mouvement et Perception
163 avenue de Luminy
CP 910
13288 Marseille Cédex 09
FRANCE