Data Augmentation for Brain-Computer Interfaces: Analysis on Event-Related Potentials Data
This addresses a data scarcity issue for brain-computer interface researchers, though it is incremental as it adapts existing augmentation techniques to a new domain.
The paper tackled the problem of insufficient training data for electroencephalographic (EEG) data in brain-computer interfaces by proposing and evaluating temporal and spatial/rotational distortions for data augmentation, resulting in performance increases of 1% to 6% on P300 and MRCP tasks.
On image data, data augmentation is becoming less relevant due to the large amount of available training data and regularization techniques. Common approaches are moving windows (cropping), scaling, affine distortions, random noise, and elastic deformations. For electroencephalographic data, the lack of sufficient training data is still a major issue. We suggest and evaluate different approaches to generate augmented data using temporal and spatial/rotational distortions. Our results on the perception of rare stimuli (P300 data) and movement prediction (MRCP data) show that these approaches are feasible and can significantly increase the performance of signal processing chains for brain-computer interfaces by 1% to 6%.