"American Sign Language (ASL) fingerspelling recognition using marker-less vision sensors is a challenging task due to the complexity of ASL signs, self-occlusion of the hand, and limited resolution of the sensors. This thesis describes a new method for ASL fingerspelling recognition using a low-cost vision camera, which is Microsoft's Kinect. A segmented hand configuration is first obtained by using a depth contrast feature based per-pixel classification algorithm. Then, a hierarchical mode-finding method is developed and implemented to localize hand joint positions under kinematic constraints. Finally, a Random Decision Forest (RDF) classifier is built to recognize ASL signs according to the joint angles. To validate the performance of this method, a dataset containing 75,000 samples of 24 static ASL alphabet signs is used. The system is able to achieve a mean accuracy of 92%. We have also used a publicly available dataset from Surrey University to evaluate our method. The results have shown that our method can achieve higher accuracy in recognizing ASL alphabet signs in comparison to the previous benchmarks."--Abstract, page iii.
Leu, M. C. (Ming-Chuan)
Liou, Frank W.
Mechanical and Aerospace Engineering
M.S. in Mechanical Engineering
Missouri University of Science and Technology
vii, 39 pages
© 2015 Cao Dong, All rights reserved.
Thesis - Open Access
American Sign Language -- Computer simulation
Kinect (Programmable controller) -- Programming
Electronic OCLC #
Link to Catalog Record
Dong, Cao, "American Sign Language alphabet recognition using Microsoft Kinect" (2015). Masters Theses. 7392.