Kinect finger tracking

Creating a human-computer interaction environment which is robust and effective was difficult a few years ago, but thanks to the launch of the Kinect the technology has become less problematic. Evidently, the technology has become more user friendly thanks to the software implemented in the Kinect, allowing the user to focus more on code implementation and less on data management. The immediate objective of this project is to create a robust system that will allow effective finger tracking without extra peripherals on the users’ hands. For this summer project our first step involved finding the hand contour, for this the camera captured image is converted to grey scale the image is smoothen and threshold. Once we obtain our contour, we analyze each point in order to determine which point is a fingertip. A fingertip tip or valley is detected by using K-curvature algorithm (angle between two vectors). Vectors A(P_i, P₁) and B(P_i, P₂). Point P is indexed i, a countour point along the edge of the hand. Point P1 is i+r, where r is a set value. P2 is indexed i-r. K- values are calculated from vectors A and B. K values closer to 0 are possible candidated for peaks or valleys. Points are surveyed for valleys or peaks. The cross product of vectors A and B are taken; if the Z component of the vector is positive then it is labeled a peak, whereas a negative value is considered a valley.