Compressive sensing based facial expression recognition

In this paper a new approach is proposed. The new approach is inspired by the compressive sensing theory used for data compression. Feature vectors from the facial data were extracted using compressive sensing technique. The compressive measurements were obtained at different feature vector lengths. Comparison of recognition rates with Local Binary Pattern (LBP) feature extractor were made using different classifiers. The results obtained by using three distance measures and Support Vector Machines (SVM) validate the performance of our proposed approach as it outperforms the best result obtained using LBP algorithms at different combinations of cells and bins numbers. Moreover, performance comparison using Sparse Representation Classifier (SRC) shows that LBP can only matches the proposed method performance with a very large feature vectors size which is computationally expensive for SRC.