Monitoring system to detect fall/non-fall event utilizing frequency feature from a microwave Doppler sensor: validation of relationship between the number of template datasets and classification performance

• Published in: Artificial life and robotics Vol. 23; no. 1; pp. 152 - 159
• Main Authors: Shiba, K., Kaburagi, T., Kurihara, Y.
• Format: Journal Article
• Language: English
• Published: Tokyo Springer Japan 01.03.2018; Springer Nature B.V
• Subjects: Artificial Intelligence; Classification; Computation by Abstract Devices; Computer Science; Control; Datasets; Gift shops; Mechatronics; Monitoring; Older people; Original Article; Robotics; Frequency feature; Template selection; Fall detection; Dynamic time warping (DTW); Microwave Doppler sensor
• ISSN: 1433-5298; 1614-7456
• DOI: 10.1007/s10015-017-0409-7

A fall event is a serious issue for the elderly because it may cause critical aftereffects. To reduce the risk of these aftereffects, early detection of the fall event is essential. However, it is difficult for caregivers to detect fall events early themselves, because they are required to constantly monitor the elderly to confirm their safety. Therefore, an automatic monitoring system which could detect fall events early is helpful in the healthcare field. We have proposed a fall event detection system utilizing a microwave Doppler sensor. The frequency feature is calculated, and compared with known fall or non-fall event data. However, for real-time detection, the number of template datasets must be as low as possible while maintaining high performance of the classification. In this paper, we attempt to identify the relationship between the number of template datasets and the performance of the proposed system.