Use this URL to cite or link to this record in EThOS: https://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.520572
Title: Hand tracking with parametric skin modelling using particle filter framework for gesture recognition
Author: Ongkittikul, Surachai
ISNI:       0000 0004 2691 2390
Awarding Body: University of Surrey
Current Institution: University of Surrey
Date of Award: 2010
Availability of Full Text:
Access from EThOS:
Access from Institution:
Abstract:
As the computer becomes more pervasive in modem life, users require a convenient, natural, and efficient interaction between human and machines. In the past few years, there has been much research to try to increase the capability of Human Computer Interaction (HCI) approaches. Vision-based approaches use human body parts, such as head, eyes, mouse and hands as interface devices. Hand gestures are an example of human-machine interactions that are comfortable to the user. Hand-tracking is a vital part of hand gesture recognition. There is much research to improve hand-tracking using various techniques. This thesis describes hand tracking techniques developed based on the particle filter framework. The problem under consideration is the accurate capture of human hand motions in video sequences. To clarify the target model, skin detection based on parametric skin modelling is employed as the main feature of hand tracking. From skin training states, elliptical boundaries are used to set the skin characteristics. Improvements in tracking have been demonstrated by comparing with a generic technique, such as a mean shift tracker, as well as ground truth positions. The results show that skin classification with elliptical boundary modelling can augment the performance of hand tracking to cope with rapid hand movements. The particle filter has been used as the basis for the tracking algorithm. Several features are employed for hand tracking which include skin colour and motion vector. In addition, the changing of the mechanism in the particle filter has been established, such as embedding with k-mean clustering and mean shift vector. All of the developed schemes are evaluated by comparing with the ground truth positions of objects in term of accuracy. To evaluate the tracking scheme, the framework of evaluation was employed with a variety of users and many videos to show the stability of each algorithm. The performance in this approach is better than the auxiliary particle filter and means shift iteration. The proposed approach is used in an application, namely a cash machine, which makes use of advanced HCI approaches. Finally, the concluding chapter analyses and discusses the advantages and disadvantages of each developed scheme. Future trends in hand tracking research will also be considered.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.520572  DOI: Not available
Share: