LAUSR

Brain–computer interfaces are an emerging field of medical technology that enable users to control external digital devices via brain activity. Steady-state evoked potential is a type of electroencephalogram signal that is widely used for brain–computer interface applications. Collecting electroencephalogram data is an effort-intensive task that requires technical expertise, specialised equipment, and ethical considerations. This work proposes a class-conditioned Wasserstein generative adversarial network with a gradient penalty loss for electroencephalogram data generation. Electroencephalogram data were recorded via a g.tec HiAmp using 5, 6, 7.5, and 10 Hz flashing video stimuli. The resulting model replicates the key steady-state-evoked potential features after training for 100 epochs with 25 batches of 4 s steady-state-evoked potential data. This creates a model that mimics brain activity, producing a type of symmetry between the brain’s visual reaction to frequency-based stimuli as measured by electroencephalogram and the model output.