Homepage Inria website

Section: New Results

Human Motions in VR

To carry-out natural and efficient interactions with a digital world, it is firstly necessary to recognize and evaluate the action of the user. We consequently initiated a collaboration with the Intuidoc IRISA team to adapt methods previously used in 2D gesture recognition to 3D motion. With the increasing use of head mounted display devices (especially cheap devices recently spread in the large public), the problem of avatar simulation and embodiment has become an important challenge. In this context, we initiated collaborative works with Hybrid to better understand embodiment and consequently imagine the future generation of avatars. Concurrently, we continued to explore the use of such technology in various application domains where human performance is a key point, such as ergonomics.

Motion recognition and classification

Participants : Franck Multon, Richard Kulpa, Yacine Boulahia.

Action recognition based on human skeleton structure represents nowadays a prospering research field. This is mainly due to the recent advances in terms of capture technologies and skeleton extraction algorithms. In this context, we observed that 3D skeleton-based actions share several properties with handwritten symbols since they both result from a human performance. We accordingly hypothesize that the action recognition problem can take advantage of trial and error approaches already carried out on handwritten patterns. Therefore, inspired by one of the most efficient and compact handwriting feature-set, we proposed a skeleton descriptor referred to as Handwriting-Inspired Features [20]. First of all, joint trajectories are preprocessed in order to handle the variability among actor’s morphologies. Then we extract the HIF3D features from the processed joint locations according to a time partitioning scheme so as to additionally encode the temporal information over the sequence. Finally, we used Support Vector Machine (SVM) for classification. Evaluations conducted on two challenging datasets, namely HDM05 and UTKinect, testify the soundness of our approach as the obtained results outperform the state-of-the-art algorithms that rely on skeleton data.

This work has been carried-out in collaboration with the IRISA Intuidoc team, with Yacine Boulahia who is a co-supervised PhD student with Eric Anquetil.

Avatar Embodiment in Virtual Reality

Participant : Ludovic Hoyet.

With the massive development of virtual reality products investigated by major industrial companies (Google, Facebook, HTC, Sony, etc), there is a new need for understanding what makes users immersed in virtual environments, especially regarding their relation to their virtual representation (i.e., avatar). Amongst others, an important factor is for users to feel incarnated in their avatar, which is called virtual embodiment. As more and more technological limitations are now being unlocked, understanding such factors become important to lever new immersive applications, e.g., in education, ergonomics or entertainment.

In collaboration with the EPI Hybrid (Ferran Argelaguet and Anatole Lécuyer), we explore the capacity of avatars to convey such a sense of “virtual embodiment”, i.e., the extent to which we accept an avatar to be our representation in the virtual environment. The question of embodiment originates from the famous Rubber Hand Illusion experiment of Botvinick and Cohen (1998). This experiment demonstrated that when participants are presented with a fake rubber hand positioned beside their real hidden hand, and that both hands are synchronously stroked by an experimenter, after some time participants consider their real hand to be positioned at the location of the fake rubber hand. Today, understanding how similar phenomena happen in virtual environments is crucial to provide a maximum immersion for users. For instance, previous work demonstrated that racial biases can be reduced when users are incarnated in virtual characters of a different race, or explored body weight perception by altering the morphology of the avatar. The innovative aspect of our contributions is that we explore this embodiment effect in terms of interactions of the user with the virtual environment.

So far, we explored how people appropriate avatars by evaluating how they accept different representations of their virtual hand in virtual environments. Using various representations ranging from simplistic to highly realistic when interacting in virtual environments [19], we demonstrated that the sense of ownership (i.e., the impression that the virtual hand is actually our own hand) is increased when displaying highly realistic hand representations, but that the sense of agency (i.e., the impression to be able to control the actions of the virtual hand) is stronger for less realistic representations. With the potential of VR to alter and control avatars in different ways, e.g., the user representation, we also explored how structural differences of the hand representation can influence embodiment through controlling a six-digit virtual hand [10]. We found that participants responded positively to the possibility of controlling the virtual hand despite the structural difference, and accepted it as their own to some extent. Overall, results from such experiments further our understanding of the capacity of avatars to elicit a sense of embodiment in the users, and help to design more immersive VR experiences.

VR and Ergonomics

Participants : Charles Pontonnier, Georges Dumont, Pierre Plantard, Franck Multon.

The use of virtual reality tools for ergonomics applications is a very important challenge in order to generalize the use of such devices for the design of workstations.

We proposed a framework for collaborative ergonomic design in virtual environments. The framework consists in defining design modes and metaphors that help the users (engineers, ergonomists, end-users) to find a good trade-off between their own design constraints that can be contradictory at some point. We evaluated the framework and concluded that the active user has to be carefully chosen with regard to the design specifications, since the active user is favouring systematically its own constraints. This work has been published in the Journal on Multimodal User Interfaces [14].