Hybrid Feature Extractor Using Discrete Wavelet Transform and Histogram of Oriented Gradient on...

Area of Science:

• Computer Science
• Biometrics
• Image Processing

Background:

• Biometric recognition is crucial for security and attendance.
• Palm vein biometrics offer enhanced security due to their intrinsic nature.
• Infrared palm vein images present challenges like nonuniform illumination and low contrast.

Purpose of the Study:

• To develop an accurate and reliable palm vein recognition method.
• To address the limitations of low contrast and nonuniform illumination in palm vein images.
• To evaluate the proposed method's performance using key biometric metrics.

Main Methods:

• A convolutional neural network (CNN) model named VeinCNN was developed.
• Hybrid feature extraction using Discrete Wavelet Transform (DWT) and Histogram of Oriented Gradient (HOG) was employed.
• The method was tested on five public datasets: CASIA, Vera, Tongji, PolyU, and PUT.

Main Results:

• The VeinCNN method demonstrated promising results in accuracy, Area Under the Curve (AUC), and Equal Error Rate (EER).
• The highest performance was achieved on the CASIA dataset, with 99.85% accuracy, 99.80% AUC, and 0.0083 EER.
• The hybrid feature extraction approach proved effective in overcoming image quality issues.

Conclusions:

• The proposed VeinCNN recognition scheme offers a robust solution for palm vein-based biometric verification.
• The hybrid DWT and HOG feature extraction effectively enhances recognition accuracy.
• VeinCNN shows significant potential for real-world applications in secure identification systems.