Automated Fall Detection in Smart Homes Using Multiple Radars and Machine Learning Classifiers.

Area of Science:

• Gerontology
• Biomedical Engineering
• Signal Processing

Background:

• Falls represent a critical health risk for the elderly population.
• Radar sensing technology is emerging as a viable method for fall detection systems.

Purpose of the Study:

• To classify fall detection using multiple radar sensors and machine learning (ML) classifiers.
• To evaluate the effectiveness of different ML models in discriminating elderly falls from specific activity sequences.

Main Methods:

• Utilized a publicly available dataset (N=15) with two fall sequences: falling from a stationary position (FandS) and falling while standing up (WandF).
• Computed Range-Time (RT), Range-Doppler (RD), and Doppler-Time (DT) maps from radar signals.
• Extracted Shannon entropy features and classified them using Random Forest (RF), Support Vector Machine (SVM), and Neural Network (NN) with leave-one-out cross-validation.

Main Results:

• The proposed approach successfully discriminated elderly falls.
• For FandS, RF, SVM, and NN achieved F1 scores of 55.48%, 53.33%, and 61.27%, respectively.
• For WandF, F1 scores reached 80.01% (RF), 76.42% (SVM), and 47.10% (NN), with corresponding Kappa coefficients of 0.55, 0.44, and -0.14 for RF and SVM.

Conclusions:

• The developed framework demonstrates potential for accurate fall detection in smart home environments.
• Multi-radar sensing combined with ML classifiers offers a promising solution for elderly fall monitoring.
• Further research may refine these methods for enhanced reliability and broader application.