PASCAL - Pattern Analysis, Statistical Modelling and Computational Learning

Natural stimuli for auditory BCI
Johannes Höhne and Michael Tangermann
(2011) Elsevier .

Abstract

With brain–computer interfaces (BCI), one can establish a communication pathway which solely relies on neural activity. Since visual BCI paradigms might not be suitable for users with vision impairments (e.g. patients suffering from neurodegenerative diseases) [1], auditory speller paradigms have recently been investigated [2,3]. For any kind of BCI paradigm based on event-related potentials (ERPs), the choice of stimuli is crucial. In the PASS2D paradigm described in [2], nine auditory stimuli with varying pitch and direction were presented in random order with the task to count a target stimulus. It was found that concentrating on one out of the nine tones is possible, but very demanding: two out of twelve subjects could not distinguish the stimuli at all and for three other subjects, the experiments could not be finished, possibly due to fatigue. In the present offline study, the use of phonemes instead of tones as auditory stimuli was investigated to reduce the workload and to improve applicability of the PASS2D paradigm. Short phonemes, such as ‘ta’, ‘to’ or ‘it’ were recorded from three speakers, resulting in nine different spoken stimuli (125ms each). These stimuli were used to drive a 9-class BCI in close analogy to [2]. In contrast to a BCI based on tones [2,3], all participants (N= 5) judged the phonemes as pleasant stimuli, that were easy to concentrate on. Although natural stimuli possess a diffuse temporal structure, the observed ERPs (N200 and P300 component) were very similar to those ERPs evoked by tones. Classification performance was almost equal for both types of stimuli. It can be concluded that natural stimuli can be used for driving a BCI. Phonemes stand out as a particularly suited type of stimulus for auditory ERP paradigms, due to their short duration and close relation to speech.

EPrint Type:Book
Project Keyword:Project Keyword UNSPECIFIED
Subjects:Brain Computer Interfaces
ID Code:9414
Deposited By:Martijn Schreuder
Deposited On:16 March 2012