Face Recognition Using Local Binary Pattern and Nearest Neighbour Classification

Face recognition system has been widely applied in daily activities such as identity authentication and human-computer interaction. This study aims to implement a human face recognition system with local binary pattern (LBP) and nearest neighbour algorithms. LBP algorithm was used as a descriptor to extract face features. For the classifier, it is proposed to use simple but powerful nearest neighbour methods. This paper used three public face datasets to evaluate the performance of the system, i.e. AT& T, JAFFE and Yale datasets. Train/test split (TTS) cross-validation method was conducted in the experiment to achieve a robust result. The result of this study demonstrated that the proposed system achieved 93% of accuracy for the AT& T dataset and 82% of accuracy for the Yale dataset due to the complexity of these datasets. However, it was able to recognise faces in the JAFFE dataset with the highest recognition rate of 100%.