Face Recognition System Based on Modified Sparse Representation
Xudong Yang, Yongna Liu · 2019
This paper proposes a face recognition method based on modified sparse representation. Sparse representation is an advanced data analysis algorithm based on compressive sensing. Traditionally, the sparse representation is performed on the global dictionary formed by all the training classes. Afterwards, the classification is made based on the reconstruction errors. This method did not consider the individual representation capabilities of different classes. So, a modified sparse representation is designed in this study by conducting the sparse representation on the local dictionary formed by each training class. Then, the reconstruction error of each class is computed and compared to determine the label of the test sample. In the experiments, the AR and Yale-B face image databases are employed to investigate the performance of the proposed method. The results show its effectiveness and robustness.