Loading...
| Friend's email | |
| Your name | |
| Your email | |
| enter code | |
This page was sent successfuly
- Type of Document: M.Sc. Thesis
- Language: Farsi
- Document No: 46656 (19)
- University: Sharif University Technology
- Department: Computer Engineering
- Advisor(s): Manzuri Shalmani, Mohammad Tghi
- Abstract:
- Dynamic hand gesture is an efficient approach used in Human Computer Interaction (HCI). The goal is to recognize and analyze the different hand gestures in a recorded video. There are some associated challenges such as video noises, illumination changes, rotation variance, hand occlusion which make the process more complicated.To do the research we used five different gestures out of the Sheffield hand gesture dataset which is collected from six persons under three different backgrounds and two illumination conditions. This procedure needs image processing in order to detect and segment the important parts involved in images such as hands and analyzing the changes and motions of them. We used motion based hand detection method, which needs no tracking phase so makes the algorithm faster and solves the challenges like rotation variance, hand occlusion and different illumination conditions. Then we extracted features out of the preprocessed frames which are associated to a certain region of feature space. We attempted to classify the gestures by training five different HMMs for the five type of gestures. We fit a mixture of Gaussians for each of the HMM states using K-means to get its initial parameters. In the end we evaluated this approach on the Sheffield dataset by leave-one-out cross validation method and got 88.33 % as a total accuracy which takes approximately the half of time the common approaches need because of the eliminating the tracking phase
- Keywords:
- Hidden Markov Model ; Hand Gestures ; Human Computer Interaction (HCI) ; Hand Segmentation ; Hand Detection
Digital Object List
-
محتواي کتاب
- view
Bookmark