| Title |
One-Class SVM-based Fall Detection with Ultrasonic Array |
| Authors |
유지현(Jihyeon Yoo) ; 고진환(Jinhwan Koh) |
| DOI |
https://doi.org/10.5573/ieie.2025.62.11.107 |
| Keywords |
Fall detection; Ultrasonic sensor; Anomaly detection; Imbalanced data; Artificial intelligence |
| Abstract |
In modern society, as aging and the nuclear family structure progress, the number of elderly living alone is increasing. As a result, when a fall accident occurs, there is a high possibility that the individual will not receive appropriate assistance, which can result in severe consequences. In this study, we investigate an anomaly detection system that utilizes an ultrasonic array to detect falls. During the data collection process, signals reflected from an ultrasonic transducer array were recorded and preprocessed into 2D images suitable for training. For anomaly detection models, CNN(Convolutional Neural Network) and OC-SVM(One-Class Support Vector Machine) were compared and analyzed. The experimental results showed that even though the OC-SVM was trained solely on normal data, it achieved an accuracy of 99.64% and an F1-Score of 0.9933, demonstrating effective performance in anomaly detection. This experimentally confirms that OC-SVM is well-suited as a method to address data imbalance, suggesting the potential to enhance the practicality of ultrasonic array-based fall detection systems. |