EEG-based recognition of hand movement and its parameter.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Artificial Intelligence

Background:

• Brain-computer interfaces (BCIs) decode human intentions for device interaction, vital for medical rehabilitation and human-robot collaboration.
• Decoding motor intent from electroencephalographic (EEG) signals for motor execution (ME) is in early research stages.
• Current studies lack sufficient inter-subject classification accuracy for realistic BCI applications.

Purpose of the Study:

• To investigate EEG signal-based hand movement recognition using low-frequency time-domain information.
• To develop and evaluate a deep learning model for accurate hand movement classification from raw EEG data.

Main Methods:

• Collected EEG data from thirteen healthy volunteers performing four types of hand movements, including picking up, pushing, and directional displacement tasks.
• Employed a sliding window approach to augment the dataset and mitigate EEG signal overfitting.
• Constructed an end-to-end Convolutional Neural Network (CNN)-Bidirectional Long Short-Term Memory Network (BiLSTM) model for classification.

Main Results:

• The CNN-BiLSTM model achieved high classification accuracies: 99.14% ± 0.49% for four hand movements, 99.29% ± 0.11% for picking up, 99.23% ± 0.60% for pushing, and 98.11% ± 0.23% for directional displacement.
• Comparative analysis showed the CNN-BiLSTM model outperformed other deep learning models (LSTM, CNN, EEGNet, CNN-LSTM).

Conclusions:

• The CNN-BiLSTM model demonstrates practical accuracy for EEG-based hand movement recognition and parameter decoding.
• This research advances the feasibility of BCIs for realistic applications in rehabilitation and human-robot interaction.