Gesture Recognition Using Ambient Light

There is a growing interest in the scientific community to develop techniques for humans to communicate with the computing that is embedding into our environments. Researchers are already exploring ubiquitous modalities, such as radio frequency signals, to develop gesture recognition systems. In this paper, we explore another such modality, namely ambient light, and develop LiGest, an ambient light based gesture recognition system. The key property of LiGest is that it is agnostic to lighting conditions, position and orientation of user, and who performs the gestures. The general idea behind LiGest is that when a user performs different gestures, the shadows of the user move in unique patterns. LiGest first learns these patterns using training samples and then recognizes unknown samples by matching them with the learnt patterns. To capture these patterns, LiGest uses a grid of light sensors deployed on floor. While the general idea behind LiGest seems straightforward, it is actually very challenging to put it into practice because the intensity, size, and number of shadows of a user are not fixed and depend highly on the position and orientation of a user as well as on the intensity, position, and number of light sources. We developed a prototype of LiGest using commercially available light sensors and extensively evaluated it with the help of 20 volunteers. Our results show that LiGest achieves an average accuracy of 96.36% across all volunteers.