Covert Speech vs. Motor Imagery: a comparative study of class separability in identical environments

Abstract

In this study a single experimental protocol and analysis pipeline is used: once for MI tasks, and once for covert speech tasks. The goal of this study is not to maximizing classification accuracy; rather the main objective is to provide an identical environment for both paradigms, while identifying the most important activities related to the most class dependent features. Four volunteers participated in this experiment. With four classes, the average classification accuracy for covert speech tasks is 82.5%, and for motor imagery is 77.2%. The average performance is significantly higher than chance level for both paradigms, suggesting that the results are meaningful, despite being imperfect. For motor imagery tasks the most important activities are the execution of imagined movements, and goal driven executive control for suppression of overt movements, which also occur for covert speech tasks. However, the most important activity for covert speech tasks is the linguistic processing stages of word production prior to articulation, which does not occur in motor imagery. These high-Gamma linguistic processes are extremely class dependent, which contribute to the higher performance of covert speech tasks, compared to motor imagery in an otherwise identical environment.