Recurrence quantification analysis of EEG signals for tactile roughness discrimination

Roughness recognition is an important function in the nervous system that facilitates our interactions with the environment. Previous studies have focused on the neuro-cognitive aspects and frequency-based changes in response to the roughness stimuli. In this study, we investigate the effect of different roughness levels on the nonlinear characteristics of EEG signals. Nine healthy subjects participated in the current research and touched three surfaces with different levels of roughness in a passive dynamical way. The experiment was repeated for both hands separately. During the experiment, the EEG signals were recorded. Next, three nonlinear features were extracted using the recurrence quantification analysis (RQA) method; and four classifiers were hired to distinguish six conditions, including three levels of roughness and the touching hand. The results showed that EEG nonlinear characteristics were significantly affected by the variation of surface roughness. The effects were different between touching by the left or the right hand. Moreover, it was observed that employing the RQA-based features leads to the higher accuracy of classification compared to the conventional frequency-based features. Additionally, we found that the brain representation of tactile roughness has a pseudo-random dynamic, and the amount of roughness can influence a network of brain channels. Finally, utilizing the weighted combination of different brain channels while considering the extracted nonlinear features, the LDA classification accuracy was reached 93%. Therefore, it can be suggested that not only temporal variations of brain signals but also their spatial distribution (brain channels) are important to recognize the surface roughness.