Enhancing EEG Spellers Through Natural Language Processing

Electroencephalogram (EEG) spellers are a vital communication tool for people with severe motor impairments, allowing them to express themselves by choosing letters or symbols on a screen. However, current EEG spellers face challenges such as low speed, low accuracy, and the lack of a universal method. This paper presents a novel approach that integrates a visual EEG-based speller with a Natural Language Processing (NLP) model to overcome these limitations. We use a hybrid SSVEP-RSVP model and augment it with an NLP model to create a speller that can write meaningful sentences. We evaluate our method on a between-subject scheme with four subjects and show that it significantly improves the accuracy of the basic hybrid speller (average accuracy increase by 16%). This research could potentially transform the field of assistive technology, offering a more effective and accurate communication tool for disabled people.