An Application of Statistical Modeling for Classification of Human Motor Skill Level

Automation technology has expanded dramatically in the recent years; however human manual work is still required in many domains. To facilitate appropriate design of manual skill training and to reduce individual differences in manual performance, there is a need to assess individual skill level in advance of training exercises. Unfortunately, current motor skill assessment approaches do not provide direct indicators of human operator skill levels (e.g., high, medium, or low). Therefore, there is also a need to identify appropriate methods by which to classify novice operators based on their initial motor performance. In the present study, a statistical model was developed to classify participant motor ability level using a set of features based on kinematic parameters generated through a computerized motor test. The final model achieved a classification accuracy of ~98% using a 75/25 cross validation approach. Results verified the reliability of the motor-control test and validated the quantitative motor skill classification algorithm. Based on this work, it is expected that the algorithm and the test could be applied for design of novel manual skill training approaches to compensate for performance gaps among novice operators.