Towards a Single Trial fNIRS-based Brain-Computer Interface for Communication

Document Type

Conference Proceeding

Date of Original Version

5-16-2019

Abstract

Communication based on brain-computer interface (BCI) systems is still a challenge. Although most popular classes of BCIs heavily rely on electroencephalography (EEG), recent studies have demonstrated the feasibility of using functional near-infrared spectroscopy (fNIRS) as a reliable control signal in BCI systems. However, due to the inherent latency in hemodynamic responses, these systems are considerably slow. To address this issue, this study proposes an innovative oddball-based visio-mental task and investigates the feasibility of developing an fNIRS speller. The proposed paradigm derives its principles from the conventional oddball paradigm, which has been modified to include a set of mental arithmetic operations in the flash condition. Using statistical parametric mapping (SPM) and Pearson correlation analysis, the optimum channels and hemodynamic features were selected respectively. Linear discriminant analysis (LDA) was used to evaluate the performance of the proposed fNIRS-speller. Using 2 optimum channels, our analysis demonstrated the highest average accuracy of 78.5 ± 5.7% within 2-4 seconds of the stimulation and an average accuracy of 77.0 ± 8.9% only within the first 2 seconds. Achieving satisfactory performance while using only 2 channels and a 2-second window highlights the feasibility of developing a convenient and real-time fNIRS-speller. Such a system may have potential translational applications, particularly in users with a lack of eye gaze control.

Publication Title, e.g., Journal

International IEEE/EMBS Conference on Neural Engineering, NER

Volume

2019-March

Share

COinS