A zero-shot learning approach to the development of brain-computer interfaces for image retrieval

Area of Science:

• Neuroscience
• Computer Science
• Artificial Intelligence

Background:

• Brain decoding infers cognitive states from brain activity, offering potential for human-computer interaction.
• Current methods often require extensive training data for specific tasks.

Purpose of the Study:

• To develop a zero-shot electroencephalography (EEG)-to-image brain decoding approach.
• To enable reliable and scalable identification of viewed images from brain activity.
• To enhance real-world image retrieval applications.

Main Methods:

• Utilized state-of-the-art EEG preprocessing and feature selection.
• Mapped EEG activity to biologically inspired computer vision and linguistic models.
• Employed a zero-shot learning framework.

Main Results:

• Achieved competitive decoding accuracies on two EEG datasets.
• Demonstrated the effectiveness of the zero-shot approach for image identification.
• Showcased a method more applicable to real-world image retrieval.

Conclusions:

• The proposed zero-shot EEG-to-image decoding method is effective and scalable.
• This approach offers a promising alternative to traditional classification for brain-computer interfaces.
• Advances in brain decoding can significantly impact human-computer interaction and image retrieval.