A spatial coherence-based vision-free brain-computer interface using auditory selective attention

Abstract

The development of Brain Computer Interfaces based on Auditory Selective Attention allows patients unable to voluntarily control eye movement to interact with the interface, because such systems are independent of vision. An alternative technique suitable for Brain Computer Interface applications is Spatial Coherence, an objective method to detect the side where a subject is focusing attention. This method takes into consideration the Coherence Function and the topographic distribution of responses between EEG electrodes. In this work, we further study the Information Transfer Rate and the effects of overlapping windows in the calculations. The idea behind the overlapping is to decrease the duration of the test in order to augment the Information Transfer Rate. METHOD: EEG signals were collected from fourteen adult men aged between 19 and 28 years. Amplitude-modulated tones were used for stimulation, with 32 and 38 Hz modulation and 500 and 2000 Hz carrier frequencies, on the left and right ears, respectively. Spatial coherence was used in an online Brain Computer Interface system using auditory steady-state responses modulated by Auditory Selective Attention. RESULTS: The obtained hit rates and the Information Transfer Rate may be considered appropriate, with a maximum value of 82% and 1.89 bits/min. The better detector regarding sensitivity versus specificity can be obtained by using a 50% overlap between consecutive data windows. CONCLUSION: We conclude that the Spatial Coherence successfully detected the focus of attention, and it seemed useful as a classifier of the attention condition for vision-free Brain Computer Interface.

Improving the Classifier Performance in Motor Imagery Task Classification: What are the steps in the classification process that we should worry about?

“…In this paper we have computed five FS techniques: AUC values, F-score, IG, r pb and KW. These approaches are representative from a machine learning perspective and they have been previously used in BCI systems 7,22,23,24,25,26 . We now describe the different FS techniques ex-plored in this work.…”