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.

[1]  Andreas Christmann,et al.  Support vector machines , 2008, Data Mining and Knowledge Discovery Handbook.

[2]  Danuta Roman-Liu,et al.  Comparison of concepts in easy-to-use methods for MSD risk assessment. , 2014, Applied ergonomics.

[3]  Corinna Cortes,et al.  Support-Vector Networks , 1995, Machine Learning.

[4]  Eduardo Palermo,et al.  Validation of Inter-Subject Training for Hidden Markov Models Applied to Gait Phase Detection in Children with Cerebral Palsy , 2015, Sensors.

[5]  E. Occhipinti.  OCRA: a concise index for the assessment of exposure to repetitive movements of the upper limbs. , 1998, Ergonomics.

[6]  Alexandre Campos,et al.  Real-Time Monitoring Human Motion System for Ergonomic Posture , 2017 .

[7]  Angelo M. Sabatini,et al.  Machine Learning Methods for Classifying Human Physical Activity from On-Body Accelerometers , 2010, Sensors.

[8]  Zaccaria Del Prete,et al.  Measuring Gait Quality in Parkinson’s Disease through Real-Time Gait Phase Recognition , 2018, Sensors.

[9]  Eduardo Palermo,et al.  Gait Partitioning Methods: A Systematic Review , 2016, Sensors.

[10]  Giuseppe Di Gironimo,et al.  Towards a new monitoring system to detect illegal steps in race-walking , 2017 .

[11]  Jose Antonio Diego-Mas,et al.  Errors Using Observational Methods for Ergonomics Assessment in Real Practice , 2017, Hum. Factors.

[12]  Eduardo Palermo,et al.  A HMM distributed classifier to control robotic knee module of an active orthosis , 2015, 2015 IEEE International Conference on Rehabilitation Robotics (ICORR).

[13]  Johannes Siegrist,et al.  Perceived work stress, overcommitment, and self-reported musculoskeletal pain: Across-sectional investigation , 2002, International journal of behavioral medicine.

[14]  Alberto Villa,et al.  Monitoring the human posture in industrial environment: A feasibility study , 2018, 2018 IEEE Sensors Applications Symposium (SAS).

[15]  Jun Zhong,et al.  Unsupervised Race Walking Recognition Using Smartphone Accelerometers , 2015, KSEM.

[16]  G. Verbeke,et al.  The role of physical workload and pain related fear in the development of low back pain in young workers: evidence from the BelCoBack Study; results after one year of follow up , 2005, Occupational and Environmental Medicine.