Savitzky-Golay filter energy features-based approach to face recognition using symbolic modeling

Area of Science:

• Computer Science
• Biometrics
• Image Processing

Background:

• Face recognition systems face performance degradation due to expression variations, illumination changes, and occlusions.
• The COVID-19 pandemic is expected to have a significant long-term impact on biometric face recognition.

Purpose of the Study:

• To develop novel feature extraction techniques for robust face recognition.
• To address challenges in face recognition, including variations in expression, illumination, and occlusion.
• To propose an efficient and robust person identification system using symbolic data modeling and similarity analysis.

Main Methods:

• Introduction of two novel feature extraction techniques: Savitzky-Golay differentiator (SGD) and gradient-based Savitzky-Golay differentiator (GSGD).
• SGD and GSGD extract discriminative information from various facial regions.
• A symbolic data modeling approach and similarity analysis measure are employed for feature representation and classification.

Main Results:

• The proposed SGD and GSGD descriptors demonstrate high performance across multiple datasets.
• Optimal performance rates achieved include 96-97% on LFW, 92-96% on ORL, 100% on AR, 84-93% on IJB-A, and 87-96% on the VISA database.

Conclusions:

• The developed SGD and GSGD feature extraction techniques effectively enhance face recognition performance.
• The symbolic data modeling approach provides an efficient and robust solution for person identification.
• The proposed methods show significant promise in overcoming common challenges in face recognition systems.