LBP based face recognition system for multi-view face using single sample per person
Anand Singh Jalal, Charul Bhatnagar, Mohd Aamir Khan, Madhav Singh Solanki · 2016
Most of the traditional approaches of face recognition techniques use more than one training sample per person in training stage for feature extraction. In various face recognition systems like Aadhar-card, e-passport, drive license, safety certification, access control, and law enhancement may not have multiple samples. But these systems have only single sample per person (SSPP) in training database. So, various famous systems fail to give accurate results because in which enough sample are not available. To advance the recognition accuracy of single subject per person, this paper proposed a robust one sample image recognition algorithm. Firstly, the features extraction of the face is done by using Local Binary Pattern (LBP) which is less sensitive to scaling and illumination condition, after that classification is done. Then recognition is carried out using Euclidean distance to identify the face. Experiments are done on the Face94 data-set and our own dataset having different poses, illumination and age difference. The proposed algorithm's results shows improvement in recognition rate.