Hands free : a wearable in-air gesture recognition system

Abstract

In this thesis I present my work in developing a wearable gesture interface for mobile devices with limited sensing capabilites. The proposed method utilizes a novel form of color invariant processing to augment traditional convolutional neural networks for gesture recognition. This techniques allows the system to perform color invariant recognition avoiding traditional issues with hand segmentation. Using this method several demo applications were prototyped on a mobile phone platform to test its real-time efficacy. Through a preliminary user study we confirmed that this method allows static gesture recognition in a wide variety of mobile scenarios.