Eye Gaze-based Early Intent Prediction Utilizing CNN-LSTM.

Area of Science:

• Assistive Technology
• Human-Computer Interaction
• Neuroscience

Background:

• Accurate and timely prediction of user intention is crucial for assistive technologies for individuals with severe motor or communication impairments.
• Existing methods may lack the precision or speed required for seamless human-computer interaction in these populations.

Purpose of the Study:

• To develop and validate a novel framework for the early prediction of user intent based on eye gaze patterns.
• To enhance the capabilities of assistive technologies by enabling faster and more accurate interpretation of user goals.

Main Methods:

• A framework combining Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) was developed.
• The system analyzes spatial and temporal information from eye gaze data to identify seen objects and selection order.
• Experimental data from eight human subjects were used to test the framework's efficacy.

Main Results:

• The proposed framework achieved an average accuracy of 82.27% in early intention prediction across various tasks.
• The combination of CNN and LSTM effectively processed eye gaze data for intent recognition.
• The system demonstrated reliable performance in identifying user intentions from gaze patterns.

Conclusions:

• The developed eye gaze-based framework significantly improves the early prediction of user intent in assistive technology applications.
• The CNN-LSTM approach offers a robust and effective solution for interpreting complex gaze data.
• This technology holds promise for enhancing communication and control for users with profound motor disabilities.