By Topic

Hand gesture recognition using color and depth images enhanced with hand angular pose data

Sign In

Cookies must be enabled to login.After enabling cookies , please use refresh or reload or ctrl+f5 on the browser for the login options.

Formats Non-Member Member
$33 $13
Learn how you can qualify for the best price for this item!
Become an IEEE Member or Subscribe to
IEEE Xplore for exclusive pricing!
close button

puzzle piece

IEEE membership options for an individual and IEEE Xplore subscriptions for an organization offer the most affordable access to essential journal articles, conference papers, standards, eBooks, and eLearning courses.

Learn more about:

IEEE membership

IEEE Xplore subscriptions

3 Author(s)
Pedro Trindade ; ISR - Institute of Systems and Robotics, University of Coimbra, Portugal ; Jorge Lobo ; João P. Barreto

In this paper we propose a hand gesture recognition system that relies on color and depth images, and on a small pose sensor on the human palm. Monocular and stereo vision systems have been used for human pose and gesture recognition, but with limited scope due to limitations on texture, illumination, etc. New RGB-Depth sensors, that reply on projected light such as the Microsoft Kinect, have overcome many of those limitations. However, the point clouds for hand gestures are still in many cases noisy and partially occluded, and hand gesture recognition is not trivial. Hand gesture recognition is much more complex than full body motion, since we can have the hands in any orientation and can not assume a standing body on a ground plane. In this work we propose to add a tiny pose sensor to the human palm, with a minute accelerometer and magnetometer that combined provide 3D angular pose, to reduce the search space and have a robust and computationally light recognition method. Starting with the full depth image point cloud, segmentation can be performed by taking into account the relative depth and hand orientation, as well as skin color. Identification is then performed by matching 3D voxel occupancy against a gesture template database. Preliminary results are presented for the recognition of Portuguese Sign Language alphabet, showing the validity of the approach.

Published in:

Multisensor Fusion and Integration for Intelligent Systems (MFI), 2012 IEEE Conference on

Date of Conference:

13-15 Sept. 2012