Avatars and humanoid robots use advanced technologies and theories and are getting more and more complex. Yet, these systems still fail to be as friendly and natural as a ‘real’ human in interaction. In order to determine the real limitations and key factors one must impose to a behavioural model to maintain a human-avatar/robot interaction as natural and efficient as a human-human interaction, we propose to build and use a super-WoOZ setup.
The two objectives of the project are : <br />1- To develop a research platform to analyze in real-time human-machine interaction <br />2- To investigate human-avatar/robot interactions and to find out their limits using the developed platform. <br />
Software platform to control in real-time the movements of a robot/avatar from a confederate’s movements.
Modification (decrease/increase, delay) of specific articulatory parameters in real-time.
Behavioural and neuroimaging studies to investigate the impact of the modifications on human-avatar/robot interaction.
Real-time control of rigid head motion and gaze of avatars from a confederate’s movements
Humanoid robots and avatars will become in the next future part of our everyday life. Such sociable systems will have very important impacts and benefits on the society. If these systems were able to interact as human beings, they could help for example elderly people not to be alone and enhance their quality of life. In addition, the developed platform may be interesting to various scientific communities, for instance to investigate human-human interaction in social psychology laboratories.
• Control of speech-related facial movements of an avatar from video. (accepted) Gibert, G., Leung, Y., Stevens, C. J., Speech Communication
• Realistic Eye Model for Embodied Conversational Agents. (in preparation) Gibert, G., Stevens, C. J., ACM Transactions on Graphics
• Multimodal Speech Animation from Electromagnetic Articulography Data. (accepted ) Gibert, G., Attina, V., Tiede, M., Bundgaard-Nielsen, R., Kroos, C., Kasisopa, B., Vatikiotis-Bateson, E., Best, C. T. , in 20th European Signal Processing Conference (EUSIPCO), Bucharest, Romania, 2012
• Realistic Eye Model for Embodied Conversational Agents. (submitted) Gibert, G., Stevens, C. J. , in ACM 3rd International Symposium on Facial Analysis and Animation
Avatars and humanoid robots use advanced technologies and theories and are getting more and more complex. Yet, these systems still fail to be as friendly and natural as a ‘real’ human in interaction. In fact, these systems still stand for interfaces to a machine and not to an entity in itself. They are only the embodiment of Human Computer Interfaces (HCI). The main reason of this failure may be due to the extreme complexity of human communication. Indeed, verbal and non verbal channels of communication are involved; speech but also facial expressions, body gestures and postures are part of the interaction and provide essential cues during humans’ interaction. In the avatar and robotic community, the common approach consists of building behavioural models of interaction from theories and/or observation of real data. In fact, the implementation of these models commands the limitations of those systems as they cannot replicate the richness of human behaviours. To overcome some of these limitations, the Wizard of Oz (WoOZ) technique is often used. It consists of an experimenter (the ‘wizard’) who simulates the avatar and/or the robot’s behaviours while observing the user who is not aware of the presence of the wizard. A set of pre-defined responses is designed and given the situation during the interaction the wizard chooses the adequate one. Fluidity issues in the interaction could appear; the wizard’s responses time must comply with the user’s expectation. In addition, the avatar and/or the robot rely on pre-recorded or models of gestures that cannot replicate the richness of human gestures. In order to determine the real limitations and key factors one must impose to a behavioural model to maintain a human-avatar/robot interaction as natural and efficient as a human-human interaction, we propose to build and use a super-WoOZ setup. The originality of this project relies on the full control of the avatar and/or robot’s behaviours by a human becoming a puppeteer. The project will have two clear phases: a technological one and an experimental one. During the first phase (year 1), we are going to develop an open source research platform. This platform will consist of sensors able to track human motion in 3D, software programs to convert motion into articulatory parameters and avatars and robots mimicking accurately and in real-time a confederate’s behaviours. During the second phase (year 2 and 3), a series of behavioural and neuroimaging experiments will be run to investigate several aspects of human communication. The focus will be directed towards verbal, coverbal and non verbal channels of communication and the sense of presence. The platform will allow the manipulation/degradation of specific behaviour without modifying the rest of the interaction. Fine grain investigation will be possible with this state-of-the art platform. This project will advance fundamental research in human-avatar and human-robot interactions, as well as human-human interaction.
Monsieur Guillaume GIBERT (INSTITUT NATIONAL DE LA SANTE ET DE LA RECHERCHE MEDICALE - DELEGATION REGIONALE RHONE-ALPES AUVERGNE) – email@example.com
The author of this summary is the project coordinator, who is responsible for the content of this summary. The ANR declines any responsibility as for its contents.
INSERM U846 INSTITUT NATIONAL DE LA SANTE ET DE LA RECHERCHE MEDICALE - DELEGATION REGIONALE RHONE-ALPES AUVERGNE
Help of the ANR 500,000 euros
Beginning and duration of the scientific project: December 2011 - 36 Months