Sparse Representation-Based Face Recognition

MATLAB 9K 268 views 0 downloads 1 credits
Tags:

Resource Overview

Sparse representation-based face recognition using the ORL database, containing 40 subjects with 10 images each. The implementation randomly selects training samples while using the remaining images as test samples. The final recognition accuracy is calculated as the average of 20 independent trials to ensure statistical reliability.

Detailed Documentation

This document discusses face recognition based on sparse representation techniques. The implementation utilizes the ORL database comprising facial images from 40 distinct subjects, with 10 images per individual. The algorithm randomly partitions the dataset into training and testing subsets - a specified number of images per subject are selected for training while the remainder serve as test samples. This randomization process is repeated 20 times to mitigate selection bias, with the final recognition rate computed as the mean accuracy across all trials. Key implementation aspects include sparse coding optimization using L1-norm minimization, dictionary learning for feature representation, and classification via residual analysis. The enhanced description maintains all original technical specifications while incorporating algorithmic details about the sparse recovery approach and evaluation methodology.

You May Also Like