M3T-attention: a multi-level multi-scale temporal attention transformer for EEG hand movement trajectory decoding

Key Points

• Enhanced decoding accuracy was achieved with the M3T-attention model, improving trajectory prediction from EEG data.
• The model showed a significant increase in performance metrics, particularly in identifying complex movements.
• Observational analysis using EEG signals demonstrated the efficiency of multi-level, multi-scale temporal attention methods.
• This model may enable more precise brain-computer interface applications, suggesting further exploration in real-time settings.