Inferring Model Structures from Inertial Sensor Data in Distributed Activity Recognition Scenarios

Activity-Events-Detectors digraphs describe the relations between human activities and sensor nodes under a distributed perspective. The graphs provide a conceptual abstraction that decouples the set of activities from the sensor network with the aim of improving the recognition performances and lowering the computational constraints of the detection tasks in the sensor node. In this work, a data-driven methodology that learns groups of activities and infers the con?guration of detectors embedded in the sensors nodes of the network is proposed. The methodology, de?ned on a clustering procedure, derives and infers all the relevant information from sensors data, making no a-priori assumptions on the relations between sensor nodes and activities. Using the inferred structured models, a performance boost of 15% in the ?nal classi?cation accuracy is obtained with a signi?cant reduction of the computational resources needed for recognition purposes.