Hand detection and segmentation using smart path tracking fingers as features and expert system classifier

Nowadays, hand gesture recognition (HGR) is getting popular due to several applications such as remote based control using a hand, and security for access control. One of the major problems of HGR is the accuracy lacking hand detection and segmentation. In this paper, a new algorithm of hand detection will be presented, which works by tracking fingers smartly based on the planned path. The tracking operation is accomplished by assuming a point at the top middle of the image containing the object then this point slides few pixels down to be a reference point then branching into two slopes: left and right. On these slopes, fingers will be scanned to extract flip-numbers, which are considered as features to be classified accordingly by utilizing the expert system. Experiments were conducted using 100 images for 10-individual containing hand inside a cluttered background by using Dataset of Leap Motion and Microsoft Kinect hand acquisitions. The recorded accuracy is depended on the complexity of the Flip-Number setting, which is achieved 96%, 84% and 81% in case 6, 7 and 8 Flip-Numbers respectively, in which this result reflects a high level of finite accuracy in comparing with existing techniques