Human activity recognition based on energy efficient schemes

Human activity recognition (HAR) has been an active area of research for decades. While traditional sensor-based activity recognition methods have demonstrated high recognition accuracy, they suffer from a significant overhead in terms of energy and computation, especially for resource-constrained devices. To address that, this thesis employs a multi-faceted approach to arrive at an optimized system where the design involves optimization of energy consumption through number of sensors, computation through minimal set of features and reduced classifier size and cost through a nominal hardware platform.