Current Proceedings on Technology
Yazarlar: Rifat Edizkan, Hakan Çevikalp, Hasan Serhan Yavuz
Konular:-
Anahtar Kelimeler:Discriminative common vector,Face recognition,Embedded system
Özet: The Discriminative Common Vector (DCV) is an appearance-based method proposed for face recognition. In DCV, the modified Fisher’s Linear Discriminant criterion is maximized for an optimal solution to the small sample size case. In our study, DCV-based face recognition system is designed and implemented on the embedded platform. The system captures image by the camera and performs a series of processes: face detection, pre-processing, feature extraction and classification. The gray levels and Local Binary Patterns (LBP) are used to describe the face images. The system is tested under three different illumination conditions to evaluate the performances of used features. Experimental results show that the LBP features with DCV give better performance than the gray level values. The average classification accuracy of the system is 98.33% for the selected small-sized database. The DCV-based face recognition system gives encouraging results with respect to classification accuracy and processing time for real-time applications.