Rapid Decoding of Hand Gestures in Electrocorticography Using Recurrent Neural Networks

Area of Science:

• Neuroscience
• Biomedical Engineering
• Machine Learning

Background:

• Brain-computer interfaces (BCIs) offer rehabilitation for motor disabilities.
• Electrocorticography (ECoG) signals are rich in motor activity information.
• Existing decoders often overlook temporal dynamics in ECoG signals.

Purpose of the Study:

• To develop a robust hand gesture decoding method using ECoG signals.
• To exploit temporal information in ECoG signals for improved decoding accuracy.
• To investigate rapid gesture recognition shortly after motion onset.

Main Methods:

• Recurrent Neural Networks (RNNs) were employed to analyze ECoG time series.
• The RNN model was designed to capture nonlinear temporal dynamics.
• The method was tested for decoding three hand gestures from two participants' ECoG data.

Main Results:

• Achieved 90% accuracy in decoding three hand gestures.
• Demonstrated rapid gesture recognition within 0.5 seconds after motion onset with ~80% accuracy.
• Confirmed the significant contribution of temporal dynamics to decoding effectiveness and speed.

Conclusions:

• RNNs effectively leverage temporal dynamics in ECoG signals for robust hand gesture decoding.
• The proposed method enables rapid and accurate BCIs for motor rehabilitation.
• Temporal information is crucial for both effective and timely gesture recognition.