Elderly Daily Activities Recognition System Based On Implicit Perception

The ageing of China' s society has intensified and the number of empty nesters has soared.There is a hidden danger for the elderly who live alone,and their children are unable to obtain the health information of the elderly in time.Behavior identification monitoring is an assessment of personal health functions,which is an important step to judge whether the residents have independent living ability.In addition,effectively recognizing ongoing behaviors in a home environment can help with fall monitoring,behavioral analysis,behavioral alerts,chronic disease management and rehabilitation feedback.Existing behavior identification methods mostly use wearable devices and video to collect data which are hard for the elder to accept.What's more,few systems have long-term monitoring of the elderly to identify the decline in the health of the elderly.Therefore,an implicit cognitive-based daily activity recognition system for the elderly was designed and implemented which can identify the daily activity of the elderly and report abnormal behaviors of the elderly.Through the long-term monitoring of the elder's daily behaviors and the establishment of health records for the elderly,it is easy for the elderly and their family members to build a bridge of affection,so as to timely intervene in the care of the elderly.The work of this thesis mainly includes the following aspects:(1)The data of the system was collected by the hidden sensors such as magnetic switch and water flow sensor,and the transmission and communication of the data was provided by bluetooth.Finally,the system were implemented from the underlying hardware deployment and data communication protocol design to data collection,data processing and web platform building.(2)In the system,different functional requirements were designed for different users.For the elderly and elderly relatives,the visualization function of the elder's daily behavior recognition,and the abnormal behavior alarm function for the elderly were designed and implemented.For the medical staff,the management function and daily behavior recognition function life for the elderly were designed and implemented.For management personnel,management functions for all elderly people's information management,sensors information and abnormal information was designed and implemented.(3)A comparative study was conducted on the behavior recognition algorithm.The experiments of SVM,NB,and XGBoost were carried out using the data collected from sensors and the experimental results show that the XGBoost algorithm can better identify the daily behavior of the elderly.Sensing and recognizing the daily activities of the elderly through different types of implicit sensors is conducive to discovering the health decline of the elderly in a timely manner,so that timely intervention and active health management can be implemented.What's more,elderly people with independent living ability can realize safe and easily acceptable home monitoring and take care to maintain its good quality of life.It can effectively alleviate the lack of capacity in nursing service institutions in China and the tension of nursing staff.