Abstract:Proactively detecting falls and preventing injuries are among the primary keys to a healthy life for the elderly. Near-fall remote monitoring in daily life could provide key information to prevent future falls and obtain quantitative rehabilitation status for patients with weak balance ability. In this study, we developed a deep learning-based novel classification algorithm to precisely categorize three classes (falls, near-falls, and activities of daily living (ADLs)) using a single inertial measurement unit (IMU) device attached to the waist. A total of 34 young participants were included in this study. An IMU containing accelerometer and gyroscope sensors was fabricated to acquire acceleration and angular velocity signals. A comprehensive experiment including thirty-six types of activities (10 types of falls, 10 types of near-falls, and 16 types of ADLs) was designed based on previous studies. A modified directed acyclic graph-convolution neural network (DAG-CNN) architecture with hyperparameter optimization was proposed to predict fall, near-fall, and ADLs. Prediction results of the modified DAG-CNN structure were found to be approximately 7% more accurate than the traditional CNN structure. For the case of near-falls, the modified DAG-CNN demonstrated excellent prediction performance with accuracy of over 98% by combining gyroscope and accelerometer features. Additionally, by combining acceleration and angular velocity the trained model showed better performance than each model of acceleration and angular velocity. It is believed that information to preemptively handle the risk of falls and quantitatively evaluate the rehabilitation status of the elderly with weak balance will be provided by monitoring near-falls.

A fall detection study based on neural network algorithm using AHRS

Human Fall Detection Improvement Based on Artificial Neural Network and Optimized Zero Moment Point Algorithms

ARD: accurate and reliable fall detection with using a single wearable inertial sensor

Evaluation of a Low-Complexity Fall Detection Algorithm on Wearable Sensor Towards Falls and Fall-Alike Activities

A Fall Detection Device Based on Single Sensor Combined with Joint Features

A Research On Automatic Human Fall Detection Method Based On Wearable Inertial Force Information Acquisition System

Fall Detection Algorithm Based on Inertial Sensor and Hierarchical Decision

Deep Learning-Based Near-Fall Detection Algorithm for Fall Risk Monitoring System Using a Single Inertial Measurement Unit

HMM-Based Human Fall Detection and Prediction Method Using Tri-Axial Accelerometer

A Reliable Fall Detection System Based on Wearable Sensor and Signal Magnitude Area for Elderly Residents

Fall Detection System Using Millimeter-Wave Radar Based on Neural Network and Information Fusion

MEMS Tri-Axial Accelerometer Based Fall Detection

A Class-Imbalanced Deep Learning Fall Detection Algorithm Using Wearable Sensors

AFall: Wi-Fi-Based Device-Free Fall Detection System Using Spatial Angle of Arrival.

A Wearable Micro-Electromechanical System Inertial Sensor System for Fall Behaviour Detection Based on a Multi-Level Threshold Algorithm

Fall Detection Method based on a Human Electrostatic Field and VMD-ECANet Architecture

Application of Inertial and Magnetic Sensors in Human Fall Detection

Implementation of a real-time fall detection system based on hybrid threshold analysis algorithm and machine learning algorithm

Fall Detection: Threshold Analysis of Wrist-Worn Motion Sensor Signals

Triaxial Accelerometer-Based Falls and Activities of Daily Life Detection Using Machine Learning