Real-time Open-vocabulary Sentence Decoding from MEG Signals Using Transformers

Brain-computer interfaces (BCIs) hold immense potential for restoring communication to individuals with severe speech impairments. This paper investigates the feasibility of real-time open-vocabulary sentence decoding from magnetoencephalography (MEG) signals using deep learning with a Transformer architecture. We propose a novel end-to-end model that leverages the high temporal and spatial resolution of MEG and the powerful sequence-to-sequence learning capabilities of Transformers. Our model is trained and evaluated on a large dataset of MEG recordings paired with natural language sentences. We demonstrate the effectiveness of our approach in achieving real-time, accurate, and flexible communication, significantly outperforming existing methods. Our findings pave the way for the development of more practical and user-friendly BCIs for individuals with speech disabilities.