The Research On Applying Gesture Recognition Techniques To Human-Machine Interaction With Kinect

"Living room entertainment" requires a more natural way to interact with a smart TV. In addition to traditional TV functions, Smart TV also offers gaming, web browsing and other new features. These new capabilities need support from new interaction technology. During recent years, with the development of gesture recognition technique, and the launch of Kinect somatosensory equipment, applying Kinect based gesture recognition techniques to efficiently and easily interact with Smart TV has become a trend.This thesis focuses on the study on Smart TV oriented interaction techniques. Many studies show that dynamic gestures can be applied to daily operations on Smart TV to gain more efficiency and convenience. Therefore, the techniques and algorithms to recognize dynamic gestures is the main topic, and explored by using Kinect.This thesis first introduces the background and significance of human-machine interaction technology based on gesture recognition techniques, and describes the state of the art. The next section describes commonly used procedures and algorithms for gesture recognition, and defines a set of dynamic gestures for common operations in Smart TV interaction scenarios such as cursor navigation, objects grabbing, in-air handwriting. Then gestures are divided into 2 groups:the static ones and the dynamic ones, and discussion about the recognition procedures on both kinds of gestures are carried out. After dynamic gestures are recognized, DTW algorithm is adopted for 3D handwriting trajectory recognition. In order to improve the recognition rate, this thesis proposes the use of position-similarity weights. Experimental results show that the average recognition rate of improved DTW algorithm increases by 5%. Finally, the software and hardware environments for developing gesture recognition system and the operating system it runs on is introduced, and software architecture is described. The experiments use Kinect for Windows for gesture data acquisition, and then the system is tested.