Home // ACHI 2016, The Ninth International Conference on Advances in Computer-Human Interactions // View article

Hand Gesture Recognition Using SIFT Features on Depth Image

Authors:
Hasan Mahmud
Md. Kamrul Hasan
Abdullah-Al- Tariq
Md. Abdul Mottalib

Keywords: Hand gesture recognition, SIFT features, depth image, HCI, SVM

Abstract:
In this paper, we present a hand gesture recognition system using Scale Invariant Feature Transform (SIFT) on depth images. Due to SIFT features and depth information, our approach is robust against rotations, scaling, illumination conditions, cluttered background, and occlusions. Previously, SIFT features were applied on binary images which do not provide enough discriminating key points to classify close gestures. We have extracted SIFT keypoints from each depth silhouette and applied k-means clustering to reduce feature dimensions. Bag-of-word features were generated using vector quantization technique, which maps keypoints from each training image into a unified dimensional histogram. These bag-of-word features were fed into multiclass Support Vector Machine (SVM) classifier for training. We have tested our results for five close symbolic gestures, compared the results for both binary and depth images and found higher accuracy for depth images. The proposed recognition scheme can be used to develop human gesture based interactive Human-Computer Interaction (HCI) applications.

Pages: 359 to 365

Copyright: Copyright (c) IARIA, 2016

Publication date: April 24, 2016

Published in: conference

ISSN: 2308-4138

ISBN: 978-1-61208-468-8

Location: Venice, Italy

Dates: from April 24, 2016 to April 28, 2016