Presentation of an SMR-based BCI using auditory feedback based on pitch

Brain-Computer Interfaces (BCIs) aim to translate the users cerebral activity into a control signal, allowing him to interact with an external device without the use of conventional muscled-based pathways. These alternative communication pathways have already been employed under various forms to allow severely handicapped people to use assistive devices. Given that most of currently developed BCIs only provide a visual feedback to the user, the aim of this study was to observe the result provided by an auditory feedback based on pitch. The BCI developed to this end classifies two different commands using sensorimotor rhythm (SMR) frequency features. The experiment consisted in sessions of trials during which the subjects were asked to perform motor imagery or to rest. Except for the first session, the system provided the subjects with a realtime feedback based on the learning data of the previous sessions. This feedback was generated by a control signal resulting of the classification of the extracted features by a Support Vector Regression (SVR) algorithm. Half of the 16 subjects were given visual feedback, consisting in a cursor movement, and the other half received auditory feedback, consisting in a pitch variation. Although the performance of the visual feedback BCI was superior during the first sessions, the final third session classification rates were 69.7 and 70.3% for respectively visual and auditory conditioning. These results tend to show that there might be no significant difference between the two types of feedback after long enough training. Moreover, this study also demonstrate the usability of a SVR classifier.