Automated activity recognition of construction workers using single in-pocket smartphone and machine learning methods

Abstract Automatic recognition of construction workers’ activities contributes to improving productivity and reducing the potential risk injury. Kinematics sensors have been proved feasible efficient recognize activities. However, most need be tightly tied bodies, which might result in uncomfortableness reluctance wear sensors. To solve problem, this paper proposes a less physically intrusive method with single in-pocket smartphone. The smartphone was placed pocket natural non-fixed manner, its built-in accelerometer gyroscope collecting motion data. Machine learning-based classifiers were trained An experiment simulating rebar designed verify effectiveness proposed method. results showed that could identify (with an accuracy over 94%) non-intrusive manner.