Public thesis defense of Mehdi Ousmer

A Systematic Procedure for Comparative Testing of 3D Hand Gesture Template-Based Recognizers in Multiple Contexts of Use

The advent of embedded and wearable sensors and devices with precise tracking capabilities has paved the way for 3D gesture recognition. Designing usable gesture-based user interfaces is already challenging to address both gesture usability and recognition. This problem is even more challenging when such user interfaces need to be deployed in multiple contexts of use, i.e., for different users, devices, and environments. While the conceptualization of gesture recognition systems for multiple contexts of use, primarily multiple devices, is possible, their technological implementation is limited to a few devices, that are hard to transpose from one device to another. This limitation occurs because it is currently impossible to develop it for all types of devices due to their variety. To address these challenges, this thesis defines, and justifies a systematic procedure for comparative testing of 3D hand gesture recognition in multiple contexts of use. This procedure leads to a method for comparative testing of multi-device gesture recognition that is applied to two use cases: one use case with unipath dynamic gestures and one with multipath dynamic gestures. The applicability of this procedure is supported by the development of a software tool that extends a gesture recognition framework that demonstrates the potential of the comparative testing method and provides valuable insights for designers seeking a reliable and efficient recognizer to meet their needs.