TY - CHAP
T1 - Detection of Swallowing Events and Fluid Intake Volume Estimation From Surface Electromyography Signals
AU - Malvuccio, Carlotta
AU - Kamavuako, Ernest
PY - 2021/4/14
Y1 - 2021/4/14
N2 - Dehydration is a widespread problem in older adults. For this age group, dehydration can be fatal, however, there is a lack of accurate and valid methods to assess the level of hydration in healthcare settings. In this study, the use of surface electro-myographic (sEMG) sensors to detect active swallowing events (saliva vs liquid swallows) and to estimate fluid intake volume is investigated to corroborate the possibility of integrating these in a fluid intake tracking device. Eleven subjects (28\pm 13 years of age) were recruited, and sEMG recordings of single and continuous swallows from cups, bottles, and straws were analysed and four features were extracted. Results showed that the mean detection accuracy between noise and swallows was 99\pm1.31\% using a Fine Gaussian SVM classifier; and the mean classification accuracy between saliva and liquid swallows was 86.69\pm5.52\% using Fine KNN. Fluid volume was estimated using a feedforward artificial neural network, and results for single swallows presented a low mean RMSE value (2.01\pm 1.39\ \text{ml}), however the RMSE increased significantly for the estimation of continuous swallows (25.82\pm 26.39\ \ \text{ml}). These results suggest that while further validation is needed, especially in regards to volume estimation, the use of surface EMGs to track fluid intake seems promising.
AB - Dehydration is a widespread problem in older adults. For this age group, dehydration can be fatal, however, there is a lack of accurate and valid methods to assess the level of hydration in healthcare settings. In this study, the use of surface electro-myographic (sEMG) sensors to detect active swallowing events (saliva vs liquid swallows) and to estimate fluid intake volume is investigated to corroborate the possibility of integrating these in a fluid intake tracking device. Eleven subjects (28\pm 13 years of age) were recruited, and sEMG recordings of single and continuous swallows from cups, bottles, and straws were analysed and four features were extracted. Results showed that the mean detection accuracy between noise and swallows was 99\pm1.31\% using a Fine Gaussian SVM classifier; and the mean classification accuracy between saliva and liquid swallows was 86.69\pm5.52\% using Fine KNN. Fluid volume was estimated using a feedforward artificial neural network, and results for single swallows presented a low mean RMSE value (2.01\pm 1.39\ \text{ml}), however the RMSE increased significantly for the estimation of continuous swallows (25.82\pm 26.39\ \ \text{ml}). These results suggest that while further validation is needed, especially in regards to volume estimation, the use of surface EMGs to track fluid intake seems promising.
KW - dehydration , surface EMG , swallowing signals , saliva vs liquid , volume estimation
KW - Support vector machines , Liquids , Volume measurement , Estimation , Sensor phenomena and characterization , Feature extraction , Electromyography
UR - http://www.scopus.com/inward/record.url?scp=85104839671&partnerID=8YFLogxK
M3 - Conference paper
SN - 978-1-7281-4246-3
T3 - Proceedings - 2020 IEEE EMBS Conference on Biomedical Engineering and Sciences, IECBES 2020
SP - 245
EP - 250
BT - Proceedings - 2020 IEEE EMBS Conference on Biomedical Engineering and Sciences, IECBES 2020
PB - IEEE
ER -