Sukreep S.Dajpratham P.Nukoolkit C.Yamsaengsung S.Khajontantichaikun T.Mongkolnam P.Jaiyen S.Chongsuphajaisiddhi V.Mahidol University2023-09-102023-09-102023-01-01Proceedings of JCSSE 2023 - 20th International Joint Conference on Computer Science and Software Engineering (2023) , 391-396https://repository.li.mahidol.ac.th/handle/20.500.14594/89598As the number of elderly living alone is increasing every year, some seemingly common daily activities can potentially raise the risk of serious injuries and fatal accidents for these elderly. While falls can occur anywhere, they most often occur at home, this is especially true among the elderly. Without timely notification to medical personnel and assistance, the resulting injuries could be life-threatening. As falls are caused by many different risk factors, it is necessary to identify potential incidents and make needed changes accordingly in order to reduce the risk and prevent falls. Therefore, we propose a system using surveillance cameras to detect daily activities (e.g., bending down, sitting, standing, and walking) that potentially increase the risk of falling. Moreover, we recognize high risk factors of falls such as ones that involve using the phone while performing an activity, not paying attention to obstacles, and not holding the handrails while going upstairs or downstairs. Convolutional neural network is applied for activity classification in this work. This warning system is utilized for detecting risk factors of falls that commonly occur among the elderly, which could then be used to trigger a message and/or audible alert to designated persons such as a doctor, a caregiver, or family members for timely assistance and care.Computer ScienceConference PaperSCOPUS10.1109/JCSSE58229.2023.102021472-s2.0-85169290094