The performance of an electromyography-based deep neural network classifier for external and internal focus instructions

External focusing (directs focus to apparatus, environment or the effect of the action) enhances motor skill learning and reduces muscular activity compared to internal focusing (leads focus to the movement or the action itself). Nevertheless, in case of electromyography (EMG) based systems such as prosthetic arms, robotic controllers, and robot-assisted musculoskeletal rehabilitation, low EMG activity is not preferred. This present study was based on classifying attentional focus-based EMG signals using deep neural networks. Here DB4 and HAAR wavelet coefficients of the participants' signals were extracted to be used as inputs to the proposed DNNs. The accuracy rate results found to be for DB4 and HAAR 99.07% and 99.54%, respectively. Although in previous studies, attentional focusing types could be classified at EMG activity level using artificial neural networks, our DNN results would be more reliable to be used as alternate inputs to EMG-based control mechanisms. Besides, this study also provides the potential advantages of internal focusing especially in musculoskeletal rehabilitation sessions. Verbal instructions via internal focus might lead physicians to plan effective muscular rehabilitation treatment for patients who has suffered a stroke or a disorder of lower or upper limb extremities.