Inference of Cognitive Load When Understanding Mechanical Drawings by Electroencephalography for Skill Acquisition Interviews

Abstract

Process planning needs to be strongly standardized to achieve high-quality and efficient machining without depending on the skill level of the operator. However, it is difficult to automate process planning by acquiring the skills from interviews with skillful operators due to complicated decision-making processes. In contrast, electroencephalography (EEG) has attracted considerable attention recently to obtain brain activity as a key to revealing operators’ decision-making processes. Therefore, this study aims to infer the cognitive load of the operators by EEG when understanding mechanical drawings for skills acquisition interviews. EEG data were classified by frequency bands, such as alpha and beta. The brain activity within these frequency bands reflects the characteristics of cognitive load across tasks. The degree of alpha over beta was visualized by calculating the power spectrum ratio of the EEG data. Multiple types of preliminary tasks were prepared with different difficulty levels to infer cognitive load. Subsequently, a machine-learning model was constructed by adapting the common spatial pattern method to infer cognitive load, which varied according to the difficulty level, from the obtained timeseries EEG data. After validation of cognitive load inference, a machine-learning model was applied to the EEG data obtained during the understanding of mechanical drawings to classify the difficulty levels. The inference results demonstrated the possibility of identifying time-varying cognitive load and support the interviews to acquire machining skills.