University of Twente Student Theses

Login

User-independent recognition of dynamic hand gestures using a dataglove

Ganzeboom, Mario S. (2010) User-independent recognition of dynamic hand gestures using a dataglove.

[img]
Preview
PDF
2MB
Abstract:The goal of this thesis is the user-independent recognition of dynamic, singlehanded gestures using a dataglove. This research focuses on the technology behind the recognition of these gestures. It wants to try and find suitable technologies, compare them and research ways to optimise the use of these technologies to increase recognition performance. The dynamic hand gestures described in this research were defined by researching the tasks users do when controlling a large display surface. Nine gestures were defined to cover different aspects of navigating a user interface. Using a dataglove for hand pose data and an additional sensor for position and orientation data, data was recorded on all nine gestures from eight users in a Wizard of Oz experiment. The aim of this experiment was the simulation of a real-world environment in which users control two applications with the nine gestures. With the recorded data three types of recognisers that were suitable for the goal of this thesis were trained and tested: Discrete Hidden Markov Models (DHMMs), Continuous Hidden Markov Models (CHMMs) and Latent-Dynamic Conditional Random Fields (LDCRFs). Different data pre-processing methods and parameter tuning were applied to try and increase their recognition rate. Contrary to expectation, the DHMMs and CHMMs performed poorly and only LDCRFs were able to achieve a satisfying recognition rate on the recorded data set. Initial testing of the LDCRFs on user independence also showed satisfying recognition rates. This thesis shows that the user-independent recognition of dynamic hand gestures using a dataglove is feasible. Even though this research describes only a part of the process involved in designing hand gesture interaction, it can without a doubt play an effective role in its realisation.
Item Type:Essay (Master)
Faculty:EEMCS: Electrical Engineering, Mathematics and Computer Science
Subject:54 computer science
Programme:Human Media Interaction MSc (60030)
Link to this item:http://purl.utwente.nl/essays/59720
Export this item as:BibTeX
EndNote
HTML Citation
Reference Manager

 

Repository Staff Only: item control page