Home // ACHI 2018, The Eleventh International Conference on Advances in Computer-Human Interactions // View article

Translation of Sign Language Into Text Using Kinect for Windows v2

Authors:
Preeti Amatya
Kateryna Sergieieva
Gerrit Meixner

Keywords: Sign Language; Deutsche Gebärdensprache; DGS; German Sign Language; Dynamic Time Warping; Visual Gesture Builder

Abstract:
This paper proposes methods to recognize and translate dynamic gestures of the German Sign Language (Deutsche Gebärdensprache, DGS) into text using Microsoft Kinect for Windows v2. Two approaches were used for the gesture recognition process: sequence matching using Dynamic Time Warping algorithm and a combination of Visual Gesture Builder along with Dynamic Time Warping. For benchmarking purposes, eleven DGS gestures, which were provided by an expert user from Germany, were taken as a sample dataset. The proposed methods were compared on the basis of computation cost and accuracy of these gestures. The computation time for Dynamic Time Warping increased steadily with increasing number of gestures in the dataset whereas in case of Visual Gesture Builder with Dynamic Time Warping, the computation time remained almost constant. However, the accuracy of Visual Gesture Builder with Dynamic Time Warping was only 20.42% whereas the accuracy of Dynamic Time Warping was 65.45%. On the basis of the results, we recommend Dynamic Time Warping algorithm for small datasets and Visual Gesture Builder with Dynamic Time Warping for large datasets.

Pages: 19 to 26

Copyright: Copyright (c) IARIA, 2018

Publication date: March 25, 2018

Published in: conference

ISSN: 2308-4138

ISBN: 978-1-61208-616-3

Location: Rome, Italy

Dates: from March 25, 2018 to March 29, 2018