Toward Efficient Palmprint Feature Extraction by Learning a Single-Layer Convolution Network.

Area of Science:

• Biometrics
• Computer Vision
• Machine Learning

Background:

• Existing palmprint recognition methods often overlook inherent palmprint characteristics.
• Deep learning approaches typically require large labeled datasets and raw pixel data.

Purpose of the Study:

• To develop an efficient palmprint-specific binary feature learning method.
• To create a compact single-layer convolution network for feature extraction.
• To improve palmprint recognition by utilizing unique palmprint patterns.

Main Methods:

• Palmprint-specific information is characterized by forming two kinds of ordinal measure vectors (OMVs).
• Collaborative binary feature codes are jointly learned by projecting double OMVs into complementary feature spaces unsupervised.
• Feature projection functions are integrated into OMV extraction filters, creating cascaded convolution templates for a single-layer convolution network (SLCN).

Main Results:

• The proposed SLCN enables efficient binary feature code extraction in a single-stage convolution operation.
• The method demonstrates extensibility for feature extraction with multiple OMV types.
• Experiments on five benchmark databases show promising feature extraction efficiency for palmprint recognition.

Conclusions:

• The proposed collaborative binary feature learning method and SLCN offer an efficient and effective approach for palmprint recognition.
• This method effectively utilizes inherent palmprint characteristics, outperforming traditional methods in efficiency.
• The approach is generalizable and shows strong performance across multiple datasets.