October 16, 2012
English Comp 1001, 019
In the scholarly article, "Hand Shape Estimation Using Sequence of Multiple Viewpoint", the point the authors are getting across is that gestures are an effective means of human communication on a par with language. There are many techniques that when undergo for this experiment, trying to find the best way to estimate hand shapes. One technique was to include techniques based on 3-D shape models. Another technique was to include image-based techniques. The last technique proposed a technique of using 3-D models to create 2-D models then matching images with the 2-D models. The main objective of this research was to perform hand shape estimation for applications in gesture or sign language recognition.
The first technique used image features such as, corners, edges, and bumps and then fit them to 3-D models. This wasn't the best technique because if feature extraction failed, the hand shape estimation would be wrong.
3-D models have many degrees of freedom, so the amount of calculations required when fitting the model would increase.
The second technique uses features obtained from images, therefor the amount of calculations would decrease. There are many human interface research projects that used this technique, but this technique has added constraints concerning hand shapes or palm directions in order to obtain reliable estimations.
The third technique provides the merits of all the other techniques, and the acquisition of a 3-D shape and the shortening of the computation times, but the model image must be provided for the combinations to work.
Author Yasushi Hamada is a student member of the Department of Computer-Controlled Mechanical Systems. He graduated from the Mechanical Engineering Division/Computer-Controlled Mechanical Systems in the School of Engineering at Osaka University in 2000. He was in the doctoral...