Changes in fetal movement (FM) patterns have been extensively argued as a biomarker for fetal health in obstetrics. Several research works have indicated the potential of regular FM monitoring to indicate the impending risk of stillbirth. However, the currently available FM monitoring methods, such as ultrasonography, are not suitable for regular monitoring of FM outside of clinical environments, which represents an unmet need for a reliable FM monitoring system for at-home use. Despite recent progress in wearable sensor technologies, reliable instrumentation to monitor FM out-of-clinic remains unresolved, particularly due to the challenges of separating FMs from interfering artifacts arising from maternal activities. Efforts have been focused almost exclusively on homogenous (single) sensing and information fusion modalities, such as decoupled acoustic or accelerometer sensors. However, FM and related signal artifacts have varying power and frequency bandwidth that homogeneous sensor arrays may not capture or separate efficiently.

This dissertation introduces the design of a multi-modal wearable FM monitor combining accelerometers, acoustic sensors, and piezoelectric diaphragms to capture a broad range of FM signals in terms of vibrations of a gravid abdomen. It also presents the design of a novel kick simulator that can imitate the vibrations of a gravid abdomen to test and calibrate sensors for FM monitors at the pre-clinical stage. A novel data fusion architecture combining data-dependent thresholding and machine learning is presented to automatically detect FM and separate it from signal artifacts in real-world application environments. The performance of the monitoring system was validated using 33 hours of at-home use through concurrent recording of maternally sensed FMs, which were used as the ground truth for FMs. The data collection was performed longitudinally across the gestational age (24 – 40 weeks) of each participant to enable the analysis of changes in the device performance and FM patterns with the growth of a fetus.

The multi-modal FM monitor isolated 82% of maternally sensed FMs with a precision of 76%. The performance of the multi-modal FM monitor was superior to the performance of individual types of sensors and the FM monitors from the literature (13% improvement in F1 score compared to the literature). Consistency of performance was strongest from 32 gestational weeks onwards, which overlaps with the critical monitoring window for stillbirth prevention. Analysis of longitudinal patterns of FM provided reference values of FM parameters for the current wearable FM monitor. It is expected that this multi-modal sensor fusion approach will be a major milestone in the development of low-cost wearable FM monitors that can be used in unsupervised environments.