PrimeSense sensors, also known as "Kinect" combine color and depth tracking techniques in one product. There is a whole field of applications that is developing from this technology
Abstract
PrimeSense sensors, also known as “Kinect” combine color and depth tracking techniques in one product. There is a whole field of applications that is developing from this technology, and uses it for interactions between the user and the program using only movements and no touch.
In this project we’re using this technology to recognize movements of the user, in order to use them in the future to create choreography without an assembly of dancers. We create an environment for working with the sensor, recognize user gestures, and build an interface that shows the position and movement for each one of the users by tracking and identifying the joints for each user on each frame as calculated in the pre-processing of the sensor.
In order to work in real-time with relatively small delay, this processing is done using mostly depth tracking, and therefore more fit to recognize movements than precise position. We try to improve the tracking quality with color based tracking in KLT method.
The interface allows tracking unlimited number of users, as long as enough joints of each user are in the range of the sensors. There is an option to switch between different views of the users by performing pre-defined gestures.
KLT algorithm doesn’t suit this type of tracking, and can’t be used to improve tracking in this project.
