An Energy-Efficient Algorithm for Classification of Fall Types Using a Wearable Sensor

Abstract

Objective: To mitigate damage from falls, it is essential to provide medical attention expeditiously. Many previous studies have focused on detecting falls and have shown that falls can be accurately detected at least in a laboratory setting. However, a very few studies have classified the different types of falls. To this end, in this paper, a novel energy-efficient algorithm that can discriminate the five most common fall types was developed for wearable systems. Methods: A wearable system with an inertial measurement unit sensor was first developed. Then, our novel algorithm, temporal signal angle measurement (TSAM), was used to classify the different types of falls at various sampling frequencies, and the results were compared with those from three different machine learning algorithms. Results: The overall performance of the TSAM and that of the machine learning algorithms were similar. However, the TSAM outperformed the machine learning algorithms at frequencies in the range of 10-20 Hz. As the sampling frequency dropped from 200 to 10Hz, the accuracy of the TSAM ranged from 93.3% to 91.8%. The sensitivity and specificity ranges from 93.3% to 91.8%, and 98.3% to 97.9%, respectively for the same frequency range. Conclusion: Our algorithm can be utilized with energy-efficient wearable devices at low sampling frequencies to classify different types of falls. Significance: Our system can expedite medical assistance in emergency situations caused by falls by providing the necessary information to medical doctors or clinicians. INDEX TERMS Fall detection, fall type classification, machine learning, temporal signal angle measurement, wearable device.

Human Activity Recognition by Sequences of Skeleton Features

“…However, they have the disadvantage that the devices must be attached to the subjects’ bodies. This can be uncomfortable and is not always feasible because these types of sensors must be worn constantly and in many cases present battery and wireless connection problems [ 5 , 6 , 7 ].…”

Machine Learning for Healthcare Wearable Devices: The Big Picture

“…While the reported accuracy in most of the research done for fall detection is above 90% [ 28 – 30 ], the practicality of these techniques is still questionable as the experiments were done in a controlled environment with a limited number of participants and have the limitation of a high false alarm rate [ 77 ]. Another study to simulate fall data [ 31 ] was done to generate forward and syncope accelerometer data to form a larger dataset for fall detection training.…”

Bluetooth-Low-Energy-Based Fall Detection and Warning System for Elderly People in Nursing Homes

“…In the cloud, multiple algorithm types can be used to determine whether a person has fallen. In the current literature, a lot of research is performed on machine learning algorithms [30][31][32][33][34][35][36][37][38][39] that detect falls. However, these systems require a lot of training data, which is difficult to obtain.…”