Unveiling Muscular Engagement: Evidence of Activity in Mental Imagery and Action Observation

The cognitive processes associated with the move-ment representation in the absence of physical execution, specif-ically motor imagery (MI), or with the action observation (AO), have been extensively investigated to understand how these activities can affect the endurance and the physical strength. The observed improvements of the motor performance have been attributed not only to the modifications in the central motor program but also to changes in the peripheral neuronal and musculoskeletal systems. Despite these efforts, capturing these effects has proven challenging, with inconsistent results obtained when analyzing electromyographic activity (EMG). This study aims to explore distinctive myoelectric patterns during MI and AO tasks, contrasting them with baseline patterns associated with a resting condition. Eight healthy subjects participated in imagining and observing a fine motor task involving hand and arm. Surface EMG signals from Opponens Pollicis, Flexor Radialis Carpi and Biceps Brachialis were analyzed for both experimental conditions and compared with signals obtained during rest (REST). Features in the time and the time-frequency domains were computed and two their subsets, resulting statisti-cally different for MI and for AO, in comparison to REST, were selected. These features were then inputted into three shallow machine learning models (KNN, SVM, and LDA) to assess the consistency of muscular engagements. The KNN classifier demonstrated the best performance in terms of accuracy (87 %) for MI vs REST, utilizing a subset of features exclusively defined in the time-domain. To capture muscular activity during AO, a more comprehensive feature subset, including some time-frequency features, was required to KNN to reach an accuracy of 89%.