Automatic identification and counting of repetitive actions related to an industrial worker

This study aims at proposing an innovative experimental methodology for the automatic recognition and counting of occupational repetitive action during a work cycle of an industrial worker. Through this aim, we want to evaluate the possibility to objectivize the computation of the OCcupational Repetitive Action (OCRA) index. Two healthy subjects were enrolled in the study and they were asked to perform 11 technical actions in two different repetitions. Participants were equipped with 23 inertial sensors place on different body segments in order to evaluate angles related to trunk, shoulder, elbow and wrist in the three anatomical planes. An algorithm based on Support Vector Machines was used for the automatic recognition of the technical actions; a further algorithm was implemented for the counting of the actions after their recognition. The recognition algorithm was tested with both a subject-specific training and a standard training. Results showed an accuracy in the recognition greater than 89.5% and 86.5% for the subject-specific and standard training, respectively. As regards the action counting, algorithm showed an accuracy from 72.2% to 100% based on different actions. This preliminary study opens the possibility to validate an automatic methodology for the automatic detection and counting of repetitive actions in industrial environment based on machine-learning algorithm and inertial sensors.