| Summary: | 碩士 === 中華工學院 === 電機工程學系 === 85 === Man-machine interface is becoming more and more important and
highly challenging task in these days. Many applications can be
found in Virtual Reality or multimedia area. Glove gesture
recognition plays an essential role in these applications. Due
to some technical problems, gesture recognition has not been
achieved to a satisfactory successful rate. We implemented a
spatio-temporal based real-time gesture recognition system. The
system uses spatio-temporal dynamic feature to recognize user-
defined gestures.Gesture recognition is difficult for various
reasons. The variations include the speed of the movement of a
gesture, environment changes, quality and the way describing a
gesture for various users. The traditional approaches have been
only partially successful. In this thesis, we describe our
approach to overcome the difficulties using spatio-temporal
feature extraction and radial basis function neural network. The
system extracts spatio-temporal feature by using dynamic time
warping method. We train and recognize multi-gestures using
radial basis function neural network and orthogonal radial basis
function neural network. It effectively solves the real-time
recognition problem. The resulting system can train the defined
gesture in real time and achieve a high gesture recognition rate
(over 90%).
|
|---|
