Analysis of haptic data for sign language recognition.
Cyrus Shahabi, Leila Kaghazian, Soham Uday Mehta, Amol Ghoting, Gautam Shanbhag, Margaret L. McLaughlin · 2001
Abstract: For the past two years we have been addressing the challenges involved in managing the data generated within immersive environments. We together with many other researchers have addressed the management of obvious data types such as image, video, audio and text. However, we identified a set of less familiar data types, collectively termed as immersidata, that are specific to immersive environments. In this paper, we focus our attention on analysis of a kind of immersidata known as haptic data. We propose to analyze the haptic data acquired from CyberGlove to recognize different static hand signs automatically. The ultimate objective is to understand how to model and store haptic data in a database, for similar types of applications. We propose several techniques to analyze subtle changes in hand signs and words (a series of signs). We show that our techniques can recognize the most important features to distinguish between two letters and several preliminary experiments demonstrate more than 84.66 % accuracy in sign recognition for a 10-sign vocabulary. 1