Publication:
Investigation of data glove grasping features: sum of movement and area under curve
Investigation of data glove grasping features: sum of movement and area under curve
Date
2016
Authors
Mohd Hazwan Hafiz Mohd Ali
Journal Title
Journal ISSN
Volume Title
Publisher
Research Projects
Organizational Units
Journal Issue
Abstract
The rapid development of technologies that are emerging during this era produces the evolution of human-computer interaction (HCI). Data Glove is one of sensor technologies resultant from HCI advancement. Data Glove provides vital information of finger grasping activities for HCI by providing physical data of finger bending. Over the centuries, various prototypes of data glove have been design by researcher for HCI application. UniMAP Glove or GloveMAP is an example of data glove prototype that utilize flexible bending sensor to track fingers movement. GloveMAP is capable to provide a voltage output proportional to degree of finger bending. This information is essential in designing the HCI application. However, data acquisitions from GloveMAP need to be processed and analysed in order to effectively train the computer to recognize the finger grasping information. Thus, an experiment is design to study several feature extraction methods with the assist of supervised and unsupervised clustering. Besides that, GloveMAP voltage output will be simplified into angle information. The purpose of this research is to recognize the grasping objects by using suitable feature extraction and clustering techniques. K-means and Linear Discriminant Analysis (LDA) clustering are used along with several feature extraction techniques to obtain the objects recognition rate. Angle of slopes (𝜃 ), length of slopes (ℓ), variance (ℴ2 ), standard deviation (σ), mean (𝑥̅), median (m)and the proposed feature extraction method sum of movement (SuM) and area under curve (A) are process with the feature selection method to select the best features for the recognition process. Throughout the end of research, recognition rate for K-means and LDA clustering is compared. The experimental results show that LDA achieved over 88.4% recognition rate using SuM and A as feature, meanwhile k-means achieved over 85.0% recognition rates using SuM and A feature.
Description
Master of Science in Mechatronic Engineering
Keywords
Interactive computer systems,
Tactile sensors,
Motion control devices,
Motion detectors,
Robotics,
Human-computer interaction (HCI),
Data Golve