Dynamical system representation, generation, and recognition of basic oscillatory motion gestures, and application for the control of actuated mechanisms.

Abstract

We present a system for generation and recognition of oscillatory gestures. Based on the human use of gestures, the analysis of other gesture recognition systems, and inspired by gestures used in two representative human-to-human control areas, we consider a set of oscillatory motions and refine from them a 24 gesture lexicon. Within the scope of this dissertation, the word "gesture" is defined as a family of parametrically delimited oscillatory motions generated by humans, animals, or machines. Each gesture is modeled as a linear-in-parameters dynamical system with added geometric constraints to allow for real time gesture recognition using a small amount of processing time and memory. The linear least squares method is used to determine the parameters which represent each gesture. A gesture recognition and control architecture is developed which takes the position measures of a feature and determines which parameters in a previously defined set of predictor bins best fits the observed motion. The gesture classification is then used to create a reference trajectory to control an actuated mechanism. Experiments in a real world environment show that our system can recognize human gestures on the order of 90% of the time. The gesture lexicon is expanded to include non-linear "come here" motions. We propose extensions for use in areas such as mobile robot control and telerobotics.