Brain-computer Interface Based On Auditory Attention And Cognitive Responses

Brain-computer interface provides the paralyzed patients with a new approach to restore movement ability or communicate with the external world. At present, BCI based on visual evoked potentials has been proved highly effective. However, many locked-in patients have compromised vision or lose the control of their eye movement, which limits the usability of visual BCI. Fortunately, hearing is usually preserved for most paralyzed patients, even for those nearly locked-in, which suggests that the auditory BCI may be useful for those patients. The auditory BCI systems are in the initial stage and have many problems, which poses the necessity to explore new paradigm for auditory BCI.Using the physiological mechanism of auditory event related potentials, here we proposed a novel auditory brain-computer interface based on auditory attention and cognitive responses. The stimulus was a sequence of spoken digits with sound location at different directions and the task was to silently recognize the characteristics of target voices. This paradigm employed the combination of auditory spatial attention and active mental responses. EEG data from 12 subjects were used to investigate the spatio-temporal pattern of event-related potentials in this paradigm. Some stable and distinct ERP components were elicited by target stimulus, which can be used as salient features for BCI systems. The support vector machine was adopted to detect target in offline BCI system and achieved a superior performance than other auditory BCI studies, with an averaged accuracy of 81.67%, which suggests that proposed paradigm is promising for real applications.An online auditory BCI system was developed based on the paradigm from offline research, which can help the users voluntarily express their wishes by performing designated active mental tasks. This online system consists of portable EEG amplifier and flexible software interface. Furthermore, an adaptive algorithm was incorporated into the proposed auditory BCI paradigm and the posterior probability of each digit being the target was introduced as the discriminate function for target detection. According to the online result of 8 subjects, this auditory BCI system can achieve acceptable performance for most subjects, with an average accuracy of 76.9%.In addition, independent component analysis (ICA) and its clustering algorithm based on temporal and spatial features were employed to decompose EEG signal from proposed BCI paradigm into four components. The comparison with the classical P300 paradigm indicates that the BCI paradigm based on auditory attention and cognitive responses can enhance the brain response to the target stimulus. Finally, the control experiment of non-spatial attention paradigm explained the advantage of introducing the spatial attention in auditory BCI.