Phd Position at LABSTICC/ENIB/CERV
Subject: Cognitive Architecture for a bodily coupling between human and virtual creature.
Supervisor: Pierre De Loor
Co-supervisors: Elisabetta Bevacqua and Cédric Buche
The aim of this PhD Thesis consists in developing a cognitive architecture able to reproduce the dynamics of the coupling between two humans during a bodily interaction. Various phenomena such as synchrony, imitation and subtle behavioral variations that emerge in human interactions are taken into account. The final goal is the real-time animation of a virtual character during the interaction with a user. Different contexts (like sport, theater, education…) will be considered.
The work of this thesis is divided into two main parts. The first part aims to define and implement a well-suited architecture. The second part consists in evaluating the credibility of the autonomous virtual character and the first person experience of users while interacting with it.
This thesis will be developed within a collaborative project funded by the French National Agency. The project, called INGREDIBLE, includes 3 important French Computer Science Labs (2 CNRS Labs and 1 IRISA Lab). During this project, the applicant will work in an emulative (a very dynamic and heterogeneous) environment, including researchers, other PhD students, Engineers and Postdoc students, with backgrounds in Computer Science, Mathematics, Cognitive Science and Psychology. The successful candidate will also interact closely with Human Science researchers, who are involved in the project.
The work will be realized at the European Center for Virtual Reality (CERV), a very well-known lab, specialized in autonomous behavior modeling and human/system interaction. More precisely, the thesis will be developed within the Lab-STICC. Associated labs are LIMSI (Interaction analysis), IRISA/UBS (gesture synthesis) and the Virtualys Company.
The European Center for Virtual Reality
The CERV laboratory is part of the UMR-CNRS (French National Research Center) LAB-STICC. It consists of about 30 researchers with backgrounds in Computer Science, Mathematics, Cognitive Science and Psychology. Main topics are virtual reality, artificial autonomous behavior, natural interaction and enactive cognitive sciences. CERV will provide a rich and dynamic intellectual environment for a gesture analysis oriented postdoc.
Further information about recent research activities can be found at the Center website: www.cerv.fr
Brest
Brest is a mid-sized town of 150,000 residents nestled in the natural Brest roadstead, at the west end of Brittany. Important harbor and second French military port, Brest possesses also several Educational Institutes, like Grandes Ecoles and top-level Universities, museums and Le Quartz, a very well known National Theater.
In addition to its scientific, scholarly and cultural strengths, Brest boasts outstanding recreational activities. Its residents may have an easy access to climbing, sailing, fishing, biking and swimming. Even a simple walk along the seaside becomes a unique experience thanks to the breathtaking landscape.
More information can be found here: www.brest.fr and www.brest-metropole-tourisme.fr
Application
Interested applicants can send a CV, a statement of research interests, and a list of at least two references (with email addresses) to: deloor@enib.fr
Beginning: September 2012
Monthly net salary: 1371 €
SUMMARY OF THE INGREDIBLE PROJECT
The goal of the INGREDIBLE project is to propose a set of scientific innovations in the domain of human/virtual agent interaction. The project aims at developing a prototype of an autonomous virtual actor capable of dynamic affective bodily coupling with a human whose gestures are analyzed in real-time. The main focus is on the dynamics of the coupling of gestures. Such a dynamics is considered the key point of the feeling of the “presence of the other” and the “engagement” of the users in the virtual reality systems of the future. In human-human interactions each actor’s gesture or variation of a gesture changes the other’s behavior and vice versa. Depending on the participants’ emotional states, such a mutual influence can lead to the emergence of coordination, a gesture based dialogue (composed of initiatives and imitations) which is singular. It is such a singularity, so specific to “real” human interactions, which we want to model in the real/virtual space.
In particular, our approach will include in an original manner several observed phenomena that appear in body interactions: integration of the different channels conveying the expressiveness (body position, quality of the movement, gesture), imitation, postural convergence and synchrony. The project will focus on the body expression of emotions during an interaction. It will also focus on the dynamics of the concatenation of gestures.
The project will not tackle linguistic and verbal aspects. This choice is justified by a realistic evaluation of the complexity of the task that we aim to realize in a limited time (4 years). The real-time gesture based interaction alone is already a full-fledged domain and, as a first approach, it can be separated from the analysis of the speech. The generality of body interaction mechanisms through different contexts allows us to conduct this project without taking into account the verbal behavior and to pave the way for future studies on verbal and non-verbal interactions.
The project tends to bring together several scientific domains:
- Study and formalization of the dynamics of the body interaction between humans
- Study on the body expression and posture of emotions
- Real time gesture recognition and anticipation
- Modeling of autonomous interactive agents
- Computer generation of expressive movements and real-time adaptation
The INGREDIBLE project aims also at defining a prototypic application by considering:
- A generic approach to be applied on several animation platforms and on different scenarios.
- An evaluation of the performances of various devices (Kinect, Optitrack, Moven …) and animation platforms (MARC, AréVi/HLIB, SMR).
- Several scenarios coming from different contexts (drama, sport, education…).
- Users’ suggestions which will help us to satisfy their needs and expectations.