Study on Accelerometer based Motion Classification
| Autor: | Kun-Hong Li, 李坤鴻 |
|---|---|
| Rok vydání: | 2010 |
| Druh dokumentu: | 學位論文 ; thesis |
| Popis: | 98 In this thesis, we reported our study of using a Wii Remote to capture the acceleration trajectories of motions, where those motions are considered as gestures for recognition. The aim of this study is to investigate the use of a Wii Remote in capturing motions made by a user so that those motions can be recognized to define some control commands for control electronic devices. A direct approach of motion classification with the use of some similarity measure used in Functional Artificial Neural Networks (FANN) will be proposed first. The second issue considered in our study is the use of the functional mapping characteristic of FANN for the motion classification problem. The original functions may not be easy to handle and then if the original function can be mapped by FANN to another functions which is easy to handle, the system may have more accurate classification for motions. Since two dimensional trajectories are considered in our experiments, a way of combining those two similarity measures is proposed. This measure fusion approach is based on the fuzzy concept and Gaussian functions are used to characterize the fuzzy membership functions. Our experiments show that the recognition rate of the proposed motion classification is about 96%. The use of FANN seems not helpful. A more subtle study is needed for this issue. |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|