Neural Decoding of Chinese Sign Language With Machine Learning for Brain-Computer Interfaces.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Human-Computer Interaction

Background:

• Brain-computer interface (BCI) research often focuses on gross motor skills.
• Sign language offers rich semantic information and executable commands through complex limb movements.
• Decoding sign language from neural signals remains underexplored.

Purpose of the Study:

• To explore neural features for decoding Chinese sign language using electroencephalography (EEG).
• To investigate the feasibility of decoding sign language through both motor imagery and motor execution.
• To evaluate the effectiveness of various classifiers and feature selection methods for sign language decoding.

Main Methods:

• Twenty subjects performed Chinese sign language movements (execution and imagery).
• EEG signals were analyzed using features like mean, power spectral density, sample entropy, and brain network connectivity.
• L1 regularization was employed for feature selection, and seven classifiers were used for decoding.

Main Results:

• The best average classification accuracy reached 89.90% for motor execution and 83.40% for motor imagery.
• The study demonstrated the feasibility of decoding between different sign languages.
• Source localization identified neural circuits in the visual contact and pre-movement areas.

Conclusions:

• Decoding Chinese sign language from EEG signals is feasible with high accuracy.
• The proposed decoding strategy provides a valuable reference for future limb decoding research in BCIs.
• Neural activity in visual and pre-movement areas is crucial for sign language processing.