Intuitive 3D Motion Generation
Context and challenges
In many fields (sports, rehabilitation, virtual prototyping), there is a need to communicate about movement in an intuitive, quick and unambiguous way.
In sports, for instance, one of the biggest challenges for a coach is to efficiently convey recommendations to athletes. Indeed, efficient communication is crucial for ensuring quick and continuous progress. The ideal solution would be for the coach to intuitively create the desired 3D movement and show the results directly to the athletes.
There is however no technology that offers this capability nowadays. Consequently, communication of a specific 3D movement relies on downgraded intermediate representations such as speech, text, or video. This leads to confusion, especially regarding subtle adjustments.
Objectives
This project aims at establishing the technological foundations for using 3D movement as a direct means of communication. Specifically, we create a demonstrator that allows us to:
- intuitively generate 3D movements: by using advanced statistical models such as Principal Component Analysis (PCA) and Variational Autoencoders (VAE);
- display generated 3D movements with relevant visual cues: we design visual cues overlaid on top of 3D movements to ensure efficient information transmission between runners and coaches; additionally, we investigate the use of graphic tablets for drawing movements.
Partners and funding
HumanTech Institute at HEIA-FR
Swiss BioMotion Lab at CHUV (SBML-CHUV)
Funded by the HES-SO