Research And Realization Of Gesture Recognition System Based On AndroidTV

With the rise of Android operating system, the Market share of Smart TV developed byAndroid platform is increasing year by year. The human-computer interactive way oftraditional remote control becomes difficult to satisfy people’s interactive needs. Nowadays,the gesture recognition based on computer vision has become an important research area ofnew generation of human-computer interaction and will break the traditional concept of TVoperation mode.Based on Smart TV platform, I mainly study the vision-based gesture recognition systemin this paper, including the following aspects:(1) Gesture image acquisition. This aspect analyses limitations of the process ofcapturing color image, and the gesture recognition based on depth image is proposed. Afterstudying current Android platform driver development technique, libusb technique is adoptedand camera driver is finally ported to the embedded system hardware platform.(2) Hand detection and segmentation. After analyzing the characteristic of gesturebehavior and gesture feature of depth image, methods of hand segmentation, hand tracking,gesture recognition are proposed. In the detection and segmentation aspect, we combineintroduced-frame difference with hand contour superficial recognition to confirm initial hand.This algorithm has higher examination speed and better efficiency.(3) Hand tracking and gesture recognition. Using the depth data of hand’s surfaceobtained from depth image can nearly simulate the hand feature, the tracking method thatcombining mean shift algorithm based on depth probability model with hand contour model isproposed. This algorithm effectively overcomes the defects that traditional mean shiftalgorithm can not adapt to the dimensional change of tracking target, realizes the effectivehand motion tracking and gesture recognition.The paper studies the Android development platform characteristics, designs and realizesan embedded gesture recognition system. We can browse and operate local pictures accordingto gesture input. It achieves the expected result.