Improving fNIRS-BCI accuracy using GAN-based data augmentation

Functional near-infrared spectroscopy (fNIRS) is expected to be applied to the brain-computer interface (BCI). Since a lengthy fNIRS measurement is uncomfortable for the participant, it is difficult to obtain a sufficient amount of data to train classification models; hence, the fNIRS-BCI accuracy decreases. In this study, to improve the fNIRS-BCI accuracy, we examined an fNIRS data-augmentation method using generative adversarial networks (GANs). Using fNIRS simulation data, we evaluated whether the proposed data-augmentation method could generate artificial fNIRS data. Comparing the cerebral blood flow (CBF) and the data generated by the GANs based on the CBF, it appeared that the proposed method could generate artificial fNIRS data. Although the data generated by the proposed method generally reproduced the fNIRS-data time series, they contained an extra noise component. We also evaluated a deep neural network (DNN) trained using fNIRS simulation data; its classification accuracy doubled after the data augmentation. This result suggests that the artificial fNIRS data generated by the proposed data-augmentation method is useful for improving BCI performance.