Biometric Recognition Based on Recurrence Plot and InceptionV3 Model Using Eye Movements

Area of Science:

• Biometric Recognition
• Human-Computer Interaction
• Computer Vision

Background:

• Current biometric recognition systems underutilize the correlational features within eye movement signals.
• Novel biometric modalities are crucial for enhancing security and user authentication.

Purpose of the Study:

• To propose a novel eye movement biometric recognition model leveraging recurrence plot encoding and the InceptionV3 deep learning architecture.
• To investigate the effectiveness of recurrence plot transformation for eye movement signal feature extraction.

Main Methods:

• Eye movement signals were transformed into 2-D images using recurrence plot encoding.
• The generated images were used as input for the InceptionV3 model for biometric recognition.
• Performance was evaluated using the GazeBaseV2.0 dataset, comparing horizontal and vertical gaze position signals.

Main Results:

• The proposed model achieved a high accuracy of 96.58% ± 0.66% for biometric recognition.
• Horizontal gaze position signals yielded superior recognition performance compared to vertical signals.
• Recurrence plot encoding outperformed Markov transition fields and Gramian angular field transformations.

Conclusions:

• The developed eye movement biometric recognition model demonstrates significant potential for accurate and robust user identification.
• Recurrence plot encoding offers a powerful method for extracting discriminative features from eye movement data.
• The InceptionV3 model effectively processes these image-based representations for high-performance biometrics.