期刊名称：Journal of Theoretical and Applied Information Technology
印刷版ISSN：1992-8645
电子版ISSN：1817-3195
出版年度：2021
卷号：99
期号：18
语种：English
出版社：Journal of Theoretical and Applied
摘要：A video surveillance system is a useful tool for observing and monitoring human activities in such a way that guarantees protection against risks and danger from within or from the immediate environment. The video surveillance system is already existing technology, which is simply a recording facility. In this study, the surveillance system can record video in real time and transmit it to an existing parameter that then feeds it to an intelligent approach to recognize the human activity in the recorded video. This model consists of a video surveillance system with feature vector of human beings and trained using Artificial Neural Network (ANN) algorithms (Normal and Abnormal). The model was then used to classify human activities such as hand waving, running, jumping, walking, boxing and other environment events. The pre-processing step uses a continuous stream of live AVI video format with a frame rate of 25/30 frame per seconds and a collective total number of frames as 600fps. This work consists of one normal scenario with four activities of dataset with a recognition rate of 98.5%, six abnormal activities from KTH dataset with a recognition rate of 90.8% and five abnormal activities from Weizmann dataset with a recognition rate of 83.2%. selected to evaluate the performance of the model in an indoor environment. The result obtained was 90.8% accurate.
