Abstract Many older adults live with one or more chronic conditions that require ongoing monitoring. At the same time, the aging population continues to grow, increasing interest in remote health management solutions, particularly for seniors who live alone and may face delayed assistance during medical emergencies. In this data article, we present a comprehensive simulated dataset designed to represent IoT-based remote health monitoring and fall detection scenarios in older adults. The dataset incorporates multimodal sensor data capturing physiological signals (heart rate, blood oxygen saturation (SpO₂), and body temperature) and motion-related measurements (three-axis acceleration and rotation). The dataset consists of raw inertial data, derived magnitudes, heart rate, and heart rate variability values, along with their timestamps from the sensors. It contains 2D and 3D human skeletons, and each record is labeled by a health condition (Normal, Hypertension, Hypotension, Fever, Hypoxia, Fall) with binary feature variables representing a label for fall detection and a label for health risk, respectively. This dataset serves as a promising benchmark for training and testing machine learning methods, including support vector machines (SVM), random forests, gradient boosting, and logistic regression, to automatically classify health status and critical event detection. The dataset is intended to support benchmarking and comparative evaluation of machine learning methods for health status and critical event classification.
Islam et al. (Sun,) studied this question.