The Head Posture System Based on 3 Inertial Sensors and Machine Learning Models: Offline Analyze

Abstract

The current paper proposes and presents a new wearable system for head posture recognition, based on three inertial sensors used to prevent inadequate head posture during different office daily activities. During this experiment, 9 daily office activities were evaluated. The proposed model distinguished between bad or good posture with a high accuracy using the inertial time series's raw data. The performance of the proposed wearable system was evaluated offline with the help of machine learning algorithms. The advantage of the proposed approach is the possibility of transmitting data through the Wi-Fi connection, portability, low cost, and high performance. During this experiment, the best classification performances it was obtained with Decision Extra Trees Classifier, that was achieved an accuracy equal to 96.78%.