Face Recognition MATLAB Code Based on Double-Weight Non-Negative Matrix Factorization

In this paper, I implement face recognition using MATLAB code based on double-weight non-negative matrix factorization. This approach combines two key concepts: double-weighting and non-negative matrix factorization (NMF). The double-weight mechanism introduces weighted constraints to enhance robustness against illumination variations and noise in facial images. The NMF algorithm decomposes the facial image matrix into basis and coefficient matrices while maintaining non-negative constraints, effectively capturing part-based facial features. The MATLAB implementation involves several key functions: - Data preprocessing: Normalizing and vectorizing face image datasets - Weight matrix initialization: Creating dual weight matrices for spatial and feature dimensions - Iterative optimization: Implementing multiplicative update rules with weight constraints - Feature extraction: Decomposing input images into discriminative components - Classification: Using cosine similarity or Euclidean distance for recognition Through matrix decomposition of facial images, we can extract distinctive facial features and perform accurate identification. The double-weight NMF method demonstrates particular effectiveness in handling occlusions and expression variations by assigning adaptive weights to different facial regions. This approach provides an efficient solution for face recognition tasks with improved accuracy compared to standard NMF methods. The MATLAB code includes visualization functions to display the learned basis images and weight distributions, facilitating better understanding of the feature extraction process.