Authors:
Xiaolin Sang
1
;
Shin'ichi Warisawa
2
;
Hao Zhang
1
;
Katsumi Abe
3
;
Masahiro Kubo
3
;
Kenichiro Tsuda
3
and
Ichiro Yamada
2
Affiliations:
1
The University of Tokyo, Japan
;
2
The University of Tokyo, The University of Tokyo and School of Engineering & Graduate School of Frontier Science, Japan
;
3
NEC Corporation, Japan
Keyword(s):
Drug Administration, Wearable Sensing, Swallowing Sounds, Wavelet Transform, Classification.
Related
Ontology
Subjects/Areas/Topics:
Biomedical Engineering
;
Biomedical Signal Processing
;
Devices
;
Health Information Systems
;
Healthcare Management Systems
;
Human-Computer Interaction
;
Pattern Recognition and Machine Learning
;
Physiological Computing Systems
;
Wearable Sensors and Systems
Abstract:
In recent years, chronic diseases have become the main causes of death around the world, and medication non-adherence among patients with chronic diseases is a common problem. A system for detecting drug administration behavior in daily life is strongly required. Currently, there is not a system for detecting this behavior by using wearable sensors. In this paper, we propose a wearable sensing method for detecting drug administration behavior in daily life by using swallowing sound, which is available and suitable for daily monitoring. To recognize the behavior from swallowing activities, a classification methodology using wavelet based features as feature vectors and artificial neural network as classifier is proposed. A high classification accuracy of 85.4% was achieved in classifying two swallowing activities of drinking water and taking a capsule with water. Furthermore, we also propose a compensation method for time-dependent change based on the frequency characteristics of swal
lowing sound.
(More)