Abstract
This project presents the development of a gait recognition system using TinyMachine Learning (Tiny ML) and Inertial Measurement Unit (IMU) sensors. Thesystem leverages the XIAO-nRF52840 Sense microcontroller and the LSM6DS3 IMUsensor to capture motion data, including acceleration and angular velocity,from four distinct activities: walking, stationary, going upstairs, and goingdownstairs. The data collected is processed through Edge Impulse, an edge AIplatform, which enables the training of machine learning models that can bedeployed directly onto the microcontroller for real-time activityclassification.The data preprocessing step involves extracting relevantfeatures from the raw sensor data using techniques such as sliding windows anddata normalization, followed by training a Deep Neural Network (DNN) classifierfor activity recognition. The model achieves over 80% accuracy on a testdataset, demonstrating its ability to classify the four activities effectively.Additionally, the platform enables anomaly detection, further enhancing therobustness of the system. The integration of Tiny ML ensures low-poweroperation, making it suitable for battery-powered or energy-harvesting devices.