Abstract:
"Monitoring theihealth of the foetusithroughout pregnancy requires beingiable to detect movements of the foetus. However, sinceiit can be difficult to analyseisensor dataiand differentiate between signals from the surroundings and the foetus, effectivelyiidentifying these movementsicreates aisignificant barrier. Byicreating a reliable and preciseialgorithm that can evaluate accelerometer data fromiwearable devices to identify actual foetal movements, thisiresearch addressedithe issue of foetal movement identification.
To solve the problem, ithis paper employed Arduino device based on accelerometers, also the advanced signal processing and machine learning techniques. The algorithm follows several steps toiidentify foetal movement (FM) signals. First pre-processesithe raw accelerometer data to remove noiseiusing Kalman filtering. Then artefact signal identification using thresholds and magnitude differences. Next feature dictionaryilearning methods are applied toiextract relevant features from the pre-processed data. Lasso, orthogonalimatching pursuit, adaptive filtering, iand random forest classifier is used to identify FM signals. Finally mask fusion and classificationialgorithm is carried out to identify FM signals collected from each accelerometer to identifyifinal result.
Studies are carried out to evaluate the system's performance using datasets that are iavailable to theipublic and self-builtidatasets usingitri-axis accelerometer sensors. There is a lot of potential forithe suggested approach to contribute to improved maternal and foetal healthcare."
