End-to-end P300 BCI using Bayesian accumulation of Riemannian probabilities

In brain-computer interfaces (BCI), most of the approaches based on event-related potential (ERP) focus detection P300, aiming for single trial classification a speller task. While this is an important objective, existing P300 BCI still require several repetitions to achieve correct accuracy. Signal processing and machine learning advances in mostly revolve around part, leaving character out scope. To reduce number while maintaining good classification, it critical embrace full problem. We introduce end-to-end pipeline, starting from feature extraction, composed ERP-level using probabilistic Riemannian MDM which feeds character-level Bayesian accumulation confidence across trials. Whereas only increase when flashed, our new called probabilities (ASAP), update each after flash. provide proper derivation theoretical reformulation approach seamless information signal characters. demonstrate that performs significantly better than standard methods public datasets.