Face Recognition Algorithm Based on Correlation Coefficient and Ensemble-Augmented Sparsity

Abstract

The representation-based classification method has become a research hotspot in recent years. Representation-based classifiers assign class labels directly to test samples based on a structured dictionary. The structured dictionary is composed of training samples. We call the samples in the dictionary as atoms. To further improve the expression ability of different training atoms in the classifier to the test samples, we propose an ensemble-enhanced sparse classification algorithm based on the correlation coefficient. The model proceeds from two aspects: the representation coefficient level and the similarity between different samples. First, a sparse dictionary is combined with the multiplicative property of training samples to build and solve the sparse and collaborative representation algorithm. The fusion representation coefficients are obtained by weighted sparse representation coefficients and collaborative representation coefficients. The test samples are reconstructed by fusing the representation coefficients, and the minimization recovery residuals for each class of samples are calculated. Second, the correlation coefficient value between the test sample and the training sample is calculated, and the maximum correlation quotient between the test sample and the training sample is obtained. Finally, the maximum correlation quotient and the minimization recovery residual are weighted, and the decision classification is carried out in the decision function to achieve the final face recognition. Experiments on AR, Extended Yale B, Georgia Tech and other general face databases show the effectiveness of the algorithm. The main contribution of this model is that it is more robust than a single sparse or collaborative representation model and can improve recognition accuracy.

Comparative study on the performance of face recognition algorithms

“…The face images in the database also are converted to vectors. Finally, similarity measure algorithms are used such as cosine similarity [29], correlation coefficient [30], and so on, to compare input vectors to vectors in the database. The output of similarity measure algorithms is called distance, if the distance is smaller, the faces are similar.…”

Discriminative Dictionary Learning based on Statistical Methods

“…Quantification of this association involves computing correlation coefficient ranging between [−1, 1]. In [44], Pearson product moment correlation coefficient is combined with the sparse reconstruction error of samples for face recognition. While reconstruction error tries to reduce the error between test sample and same class samples, Pearson correlation coefficient maximizes the error between test sample and other class samples, for improved classification results.…”