Using distributed accelerometers for gesture recognition and visualization

Abstract

Acceleration information captured from the hand can provide valuable information of its 3D angular pose. From this we can recognize hand gestures and visualize them. The applications for this technology range from industrial toutchless human-machine interface to hearing impairment where one uses gestures to communicate with another person. In this case it will be possible to have a translation of the gesture to another recognizable form such as a written language. The use of accelerometer sensors in the hand implies the user to have them somehow worn. The development of silicon chip manufacture allowed these sensors to fit in the top of a nail or implanted in the skin and still wirelessly communicate to a processing unit. This work demonstrates that it is possible to have gesture recognition from a clutter-free system by wearing very small devices and have them translated by a nearby processing unit. This work will focus on the processing of the acceleration information. It will be shown the use of methods to estimate hand pose, finger joints’ position, and from that recognize gestures. In this thesis it is also presented a visualization of the angular pose of the hand. This visualization can show a single render of the pose of a recognized gesture or it can provide a simple real-time (low-latency) rendering of the hand pose. The processing of the acceleration information will use the gravity acceleration as a vertical reference. This will be the core of this thesis work for the recognition and visualization of human gestures.