This project has working code example of hand detection and gesture understanding work done during my master thesis.
The project works with Kinect camera and uses color and depth image segmentations techniques to identify both hand's center of palm position in 3D space and saves these points for later gesture detection part.
Saved 3D points were used to identify basic gestures from 0 to 9 digits and shape identifications like square, rectangle, triangle or circle. However for final SVM classification, only 2D information was used in order to create feature vectors for each trajectories. Training data was recorded during the project work which is also part of the code! The prediction was measured by preconditions identified for each individual gesture and the classification/identification part was backed by SVM classifier.
prerequisite
- OpenCV_2.x
- OpenNI_2.x
- KinectSDK_1.8