Fall Detection With UWB Radars and CNN-LSTM Architecture.

Area of Science:

• Engineering
• Computer Science
• Gerontology

Background:

• Falls pose a significant health risk, potentially leading to severe injuries, prolonged suffering, and death.
• Efficient fall detection systems are crucial for timely medical intervention and care, especially for vulnerable populations.

Purpose of the Study:

• To develop and evaluate a deep neural network model for accurate fall detection in a real-world apartment setting.
• To differentiate between fall and non-fall events using ultra-wideband radar data.

Main Methods:

• Utilized three ultra-wideband radars to collect data in a 40-square-meter apartment.
• Employed a deep neural network comprising a convolutional neural network (CNN), long-short term memory (LSTM) network, and a fully connected network.
• Trained and tested the model using simulated falls (four types) from 10 participants in three locations, employing a leave-one-subject-out validation strategy.

Main Results:

• The proposed system achieved nearly 90% accuracy in fall detection.
• The leave-one-subject-out method demonstrated the system's generalization capabilities in real-world scenarios.

Conclusions:

• The integrated radar and deep learning approach shows significant promise for effective fall detection.
• This technology can enhance safety and provide timely alerts, potentially reducing fall-related morbidity and mortality.