Light Residual Network for Human Activity Recognition using Wearable Sensor Data

Abstract

This paper addresses the problem of Human Activity Recognition (HAR) of people wearing inertial sensors using data from the UCI-HAR dataset. We propose a light residual network which obtains an F1-Score of 97.6% that outperforms previous works, while drastically reducing the number of parameters by a factor of 15, and thus the training complexity. In addition, we propose a new benchmark based on leave-one(person)-out cross-validation to standardize and unify future classifications on the same dataset, and to increase reliability and fairness in the comparisons. In addition, we propose a common bechmark for the UCI-HAR dataset based on Leaving On (Participant) Out Cross-Validation (LOOCV) to justify the robustness of the architecture proposed and providing a fairer approach for comparisons.