Research on Indoor Activity Trajectory Monitoring Model for Elderly Living Alone Based on CNN-LSTM

This paper presents an innovative neural network model, integrating Convolutional NeuralNetworks (CNN) and Long Short-Term Memory networks (LSTM), designed to longitudinally monitorand predict the health status of elderly individuals living alone. In the model architecture, CNN is taskedwith extracting spatiotemporal features related to behavioral patterns and health from daily activitytrajectory maps, while LSTM leverages these feature sequences to learn the normal behavioral patterns ofthe elderly, handling long-term dependencies in time-series data. This dual approach not only enables theidentification of immediate anomalies but also forecasts potential health issues, such as Alzheimer'sdisease. By defining exceptional behavior tags and encoding them as supervised learning objectives, themodel achieves effective prediction of abnormal states among the elderly, providing a morecomprehensive and proactive solution for their healthcare monitoring.