Tactile and bone-conduction auditory brain computer interface for vision and hearing impaired users

• Published in: Journal of neuroscience methods 2015-04, Vol.244, p.45-51
• Main Authors: Rutkowski, Tomasz M., Mori, Hiromu
• Format: Article
• Language: English
• Subjects: BCI; Bone Conduction - physiology; Brain - physiology; Brain-Computer Interfaces; Data-driven filtering; EEG; Electroencephalography; Evoked Potentials, Auditory - physiology; Humans; P300; Persons With Hearing Impairments - rehabilitation; Psychophysics; ROC Curve; Somatosensory evoked potential; Touch - physiology; Vision Disorders - rehabilitation
• ISSN: 0165-0270; 1872-678X
• DOI: 10.1016/j.jneumeth.2014.04.010

•Novel multimodal tactile and bone-conduction BCI paradigm.•EEG signal processing methods leading to improved single-trial classification by applying synchrosqueezing transform based frequency domain filtering.•Application of logistic regression based classifier for improved single trial BCI results. The paper presents a report on the recently developed BCI alternative for users suffering from impaired vision (lack of focus or eye-movements) or from the so-called “ear-blocking-syndrome” (limited hearing). We report on our recent studies of the extents to which vibrotactile stimuli delivered to the head of a user can serve as a platform for a brain computer interface (BCI) paradigm. In the proposed tactile and bone-conduction auditory BCI novel multiple head positions are used to evoke combined somatosensory and auditory (via the bone conduction effect) P300 brain responses, in order to define a multimodal tactile and bone-conduction auditory brain computer interface (tbcaBCI). In order to further remove EEG interferences and to improve P300 response classification synchrosqueezing transform (SST) is applied. SST outperforms the classical time–frequency analysis methods of the non-linear and non-stationary signals such as EEG. The proposed method is also computationally more effective comparing to the empirical mode decomposition. The SST filtering allows for online EEG preprocessing application which is essential in the case of BCI. Experimental results with healthy BCI-naive users performing online tbcaBCI, validate the paradigm, while the feasibility of the concept is illuminated through information transfer rate case studies. We present a comparison of the proposed SST-based preprocessing method, combined with a logistic regression (LR) classifier, together with classical preprocessing and LDA-based classification BCI techniques. The proposed tbcaBCI paradigm together with data-driven preprocessing methods are a step forward in robust BCI applications research.