Date of Award

5-31-2016

Document Type

Thesis (Undergraduate)

Department or Program

Department of Computer Science

First Advisor

Xia Zhou

Abstract

In this paper, we explore the feasibility of reusing ambient light to recognize human gestures. We present GestureLite, a system that provides hand gesture detection and classification using the pre-existing light in a room. We observe that in an environment with a reasonably consistent lighting scheme, a given gesture will block some light rays and leave others unobstructed, resulting in the user casting a unique shadow pattern for that movement. GestureLite captures these unique shadow patterns using a small array of light sensors. Using standard machine learning techniques, GestureLite can learn these patterns and recognize new instances of specific gestures when the user performs them. We tested GestureLite using a 10-gesture dictionary in several real-world environments and found it achieves, on average, a gesture recognition accuracy of 98%.

Comments

Originally posted in the Dartmouth College Computer Science Technical Report Series, number TR2016-797.

Download
COinS