Remote Health Monitoring and Fall Detection in Elderly People
Description
Data Description This dataset provides a simulated, multimodal collection of physiological and motion-sensor data for IoT-based remote health monitoring and fall detection in elderly individuals. It contains 612 records covering 17 synthetic subjects (36 records per subject over ~3-hour monitoring windows), with 16 attributes spanning heart rate, blood oxygen saturation (SpO₂), body temperature, tri-axial accelerometer and gyroscope readings, derived acceleration/gyroscopic magnitudes, heart rate variability, and labeled health conditions (Normal, Hypertension, Hypotension, Fever, Hypoxia, Fall) with binary fall-detection and health-risk indicators. Two CSV files are included in this record for transparency and version-history purposes: Why Version 1 was created: The original repository release contained a file generated during an intermediate stage of the data simulation pipeline. At that point in development, subject-level perturbation (the step that applies independent random noise, baseline drift, and jitter to each synthetic subject's sensor readings) had not yet been applied. As a result, this file consists of a single 36-record simulation cycle repeated identically across all 17 subjects, producing 612 total rows with only 36 unique rows and 576 exact duplicates. This file, "healthmonitoringandfalldetection.csv," has been retained in the record for transparency but should not be used for analysis. Why Version 2 is corrected: During preparation of the original submission, this intermediate file was inadvertently uploaded in place of the finalized simulation output. Version 2 corrects this by replacing/supplementing the record with "Health Monitoring and Fall detection dataset.csv," which is the finalized dataset produced after the subject-level perturbation step was completed. In this version, each subject's 36-record cycle received its own independent draw of noise, drift, and jitter (rather than sharing one fixed sequence), so all 612 rows are unique with zero exact duplicates. This corrected file reflects the dataset actually used for model training and evaluation in the associated study and is the version recommended for all future analyses and benchmarking.
Files
Institutions
- University at Albany State University of New YorkNY, Albany