| Title |
Hand Feature Enhancement and User Decision Making for CNN Hand Gesture Recognition Algorithm |
| Authors |
장창영(Chang-Young Jang) ; 김태용(Tae-Yong Kim) |
| DOI |
https://doi.org/10.5573/ieie.2020.57.2.60 |
| Keywords |
Covolutional neural network ; Kinect ; Sensor fusion ; |
| Abstract |
In this study, we implemented a system that judges user's decision and recognizes hand gesture using digital camera image and Kinect's skeleton information. In order to determine the user's intention to control the device, the hand position and the head position are used in the skeleton information, and the user's hand for device control is tracked using the skeleton coordinates and the shift method. In order to recognize the hand pose of the user, CNN (Convolutional Neural Network) was used. The image to be used for CNN learning is based on the binary image based on skin color, and the hand feature enhancement method is used to improve CNN's hand pose recognition performance. Thinning was used to eliminate hand features that were not needed for hand pose recognition, and convex hulls were used to find fingertip points and enhance handtip points. As a result, it showed very good recognition accuracy and is expected to be used in devices such as TV and air conditioner through this study. |