Background: The elderly are prone to do some abnormal behaviours, such as tumbling or stepping out of the guardians’ monitoring area. These abnormal behaviours bring enormous hidden dangers to the health of the elderly, which need to be monitored effectively in order to be dealt with in time.
Objective: Provide an approach based on wireless sensor network (WSN) and multi-layer perceptron (MLP) to establish the behaviours monitoring system for the elderly.
Method: A behaviour monitoring system based on wireless sensor network and neural network is proposed in this paper, according to the behaviour characteristics of the elderly. The system collects real-time behaviour data of the elderly by wearing a bracelet with acceleration sensors wore on their hands. And then a behaviour recognition model of the elderly is established through the MLP and the collected behaviour data. The established behaviour recognition model is used to classify and identify the five typical behaviour characteristics of the elderly, such as walking, sitting, lying, standing and tumbling. At the same time, the location information of the elderly is estimated by the centroid localization technology based on received signal strength indication (RSSI) ranging.
Results: The experiment results show that the designed system can timely acquire the behaviour characteristic parameters of the elderly, and it can accurately identify the five typical behaviours with a 100% recognition accuracy rate. And also it can timely give the warning of the abnormal behaviours of the elderly such as tumbling or walking out of the active area.
Conclusion: The proposed system in this paper can accurately identify the abnormal behaviours of elderly and timely inform the guardians. The proposed monitoring method can effectively reduce the hurt to elderly, and can improve the work efficiency of guardians. And it has its theoretical and practical value.