The anterior contralateral response improves performance in a single trial auditory oddball BMI

Abstract Auditory Brain Machine Interfaces were designed for patients with severe neurofunctional disabilities, such as those suffering with late-stage amyotrophic lateral sclerosis (ALS), who have impaired eye movements or are unable to maintain gaze preventing them from using visual strategies for communication. This study explores the three-stimulus auditory oddball paradigm with binary choices (yes/no) associated with Empirical Mode Decomposition to extract features used to train and test a Support Vector Machine classifier. Data from standard EEG channels and from the N200-anterior-contralateral (N2ac) response signal were tested. Ten healthy male subjects, age 20 to 27 years, participated in the experiment. The best performance (average classification accuracy of 87.41% and information transfer ratio of 6.48 bit/min) was achieved when features extracted from the N2ac response were added to features extracted from the EEG channels. Also, the results showed that by using target stimuli with larger frequency separation helps the subjects focus better on the desired answer.