An Intelligent Recognition Method and System of “Phubber” based on Object Detection and Pose Estimation Algorithm

The existing research works regard the recognition of “phubber” as single task of object detection or human pose recognition. Its disadvantage is that its accuracy is insufficient, and it cannot effectively solve the problems of human body behavior understanding and human body overlaps. As a result, this paper provides an intelligent “phubber” recognition method that integrates object detection and pose estimation algorithms. The basic idea is to use the human pose estimation network to predict key points of each human body in video frame image, and then comprehensively determine whether the human body is a “phubber” according to the defined body behavior judgment rules, the type of hand-held object and distance of the hand-held objects from body. The experimental results show that this method outperforms existing baseline models of object detection in terms of recognition accuracy. As well, a system for cameras-based recognizing phubber” is developed based proposed method, and a verification of phubber” recognition on campus is conducted, reaching an overall recognition accuracy of 89.7%.