Classification of the EEG Evoked by Auditory Stimuli with a Periodic Carrier Frequency Coding in Order to Be Used in BCI Systems

• Main Authors: Elham Shamsi, Zahra Shirzhiyan, Ahmadreza Keihani, Morteza Farahi, Amin Mahnam, Mohsen Reza Heydari, Amir Homayoun Jafari
• Format: Article
• Language: English
• Published: Tehran University of Medical Sciences 2016-12-01
• Series: Frontiers in Biomedical Technologies
• Subjects: Rhythm; Amplitude Modulation; Brain-Computer Interface; Classification
• Online Access: https://fbt.tums.ac.ir/index.php/fbt/article/view/139

Purpose: Many of the current brain-computer interface systems rely on the patient`s ability to control voluntary eye movements. Some diseases can lead to defects in the visual system. Due to the intactness of these patients` auditory system, researchers moved towards the auditory paradigms. Attention can modulate the power of auditory steady-state response. As a result, this response is useful in an auditory brain-computer interface. As humans intrinsically enjoy listening to rhythmic sounds, this project was carried out with the aim of extraction and classification of the EEG signal patterns in response to simple and rhythmic auditory stimuli to investigate the possibility of using the rhythmic stimuli in brain-computer interface systems. Methods: Two three-membered simple and rhythmic groups of auditory sinusoidally amplitude-modulated tones were generated as the stimuli. Corresponding EEG signals were recorded and classified by means of five-fold cross-validated naïve Bayes classifier on the basis of power spectral density at message frequencies. Results: There was no significant difference between the classification performances of the responses to each group of the stimuli. All the classification accuracies, even without any noise reduction and artifact rejection, was greater than the acceptable value for being used in brain-computer interface systems (70%). Conclusion: Like the common sinusoidally amplitude-modulated tones, the novel proposed rhythmic stimuli in this project have a promising discrimination for being used in brain-computer interface systems. In addition, Power spectral density has provided an appropriate discrimination for within- and between-subject EEG classification.