Stimulus Effects on Subject-Specific BCI Classification Training using Motor Imagery

Motor-related Brain Computer Interfaces (BCIs) comprise a valuable tool towards the development of support systems targeting impaired individuals. Due to limited motion, classification training approaches mostly incorporate motor imagery paradigms, tapping into movement planning and preparation brain activity. In this context, subjects imagine executing specific motions through a target cue, while the corresponding electrophysiological activity is subsequently analyzed for movement prediction or early detection based on onset cognitive processes. Although previous studies have implemented a wide variety of classifiers and features for training, little emphasis has been given on the external trigger modality and its impact on subject-specific classification training. On that ground, we investigated differences in BCI training performance utilizing motion tasks guided via visual and auditory stimuli, testing multiple classifiers and frequency bands. Our results showed intra-subject as well as between-subject variability based on the stimulus, supporting the employment of multisensory guidance cues within BCI training protocols for optimal classification performance.