Publication: A Portable Acoustic System for Fetal Movement Detection at Home
2
0
Issued Date
2025-01-01
Resource Type
ISSN
1530437X
eISSN
15581748
Scopus ID
2-s2.0-85208407702
Journal Title
IEEE Sensors Journal
Volume
25
Issue
1
Start Page
1478
End Page
1486
Rights Holder(s)
SCOPUS
Bibliographic Citation
IEEE Sensors Journal Vol.25 No.1 (2025) , 1478-1486
Suggested Citation
Ouypornkochagorn T., Ratanasathien L., Dankul W. A Portable Acoustic System for Fetal Movement Detection at Home. IEEE Sensors Journal Vol.25 No.1 (2025) , 1478-1486. 1486. doi:10.1109/JSEN.2024.3486710 Retrieved from: https://hdl.handle.net/20.500.14740/20147
Author(s)
Author's Affiliation
Corresponding Author(s)
Other Contributor(s)
Abstract
Fetal movement indicates the health of fetuses. Pregnant women can perceive the movement and therefore regular counting of the movement is advisable. The counting, however, requires full attention and this might be difficult to count frequently. Portable devices that can serve this purpose without maternal attention are then in demand. In this work, an electro-mechanical-acoustic system for detecting fetal movement was proposed. The system contained only one device and the device's dish was in similar size to pediatric stethoscopes. The frequency to distinguish fetal movement signal was 20-371 Hz. The spectrogram of signals was created for classifying fetal movement sounds with a frequency resolution of 3 Hz. The experiments were performed on seven pregnant women at home or the workplace by attaching the device to the abdomen. A total of 31.3 h, 4.48 h/participant, were recorded. The participants were allowed to move their bodies, change their posture, or walk during experimenting. Four classifiers to identify fetal and nonfetal movement were investigated. The environment sounds including the walking sounds were included in the classification data. A high detectivity of 91.5% was found compared to maternal perception. The detectivity increased at late gestation, but it decreased when the abdominal circumference was small or very large. The convolutional neuron network (CNN) (MLP) classifier showed the highest overall performance of 81.6% F1-score and 83.8% accuracy.