Adaptive Gesture Recognition Based on Human Physical Characteristic

In gesture recognition, the database plays an important role as the foundation for the computer to recognize the input gesture. This paper discusses about the development of database to recognize gestural motion for several groups of people. Individual database is more accurate, but for a large number of recognition subject, we cannot build the database individually because it will consume more space in the system memory. To solve this problem, we need a new kind of database that ability to serve all types of people categories but using less memory on the system. The solution is grouped databases. The group database will able to identify a gesture from any subject by firstly classify the subject belong to which group. Then recognize the input gesture by using database in the dedicated group. The database is grouped based on body height, which is divided into three different groups. Five geometrical gestures were chosen to perform by the subject. The gestural motion was captured by Motion Capture System (MOCAP). Data from x, y and z-axes that were generated by the motion captured system then analyzed, classified and stored in the gesture database. An adaptive gesture recognition is presented to select the identical database to identify an unknown gesture that insert into the system. The result shows the accuracy of recognition implemented by gesture database is achieve up to 83.7% in recognizing the gesture performed by the subject.