Fetal movement counting is a key indicator of fetal health. While ultrasound is the hospital gold standard, its use is limited to a short time due to potential tissue harm. This work presents a novel wearable system for fetal movement localization using a 6-channel piezoelectric sensor array integrated into a maternal abdominal garment. A comprehensive phantom-based testbed with a robotic arm simulates fetal kicks across predefined abdominal zones. Multi-domain features were extracted from sen-sor signals and used to train various machine learning models for kick localization. The proposed system achieved a classification accuracy of 87.4 % (F1-score: 86.5 %) in localizing fetal kicks across nine distinct regions of the maternal abdomen using a multilayer perceptron. These findings demonstrate the potential of sensor-machine learning fusion for spatial fetal monitoring and early detection of complications in non-clinical environments.