An Efficient Face Detection and Recognition Method Based on Semi-Supervised Learning with Improved LPP Projection Method

Abstract

Machine learning and pattern recognition methods play a vital role in face detection and recognition from an image. Face recognition methods have an important role in various fields e.g. security, authentication and authorization. The facial recognition method extracts facial feature from a human image face. A high accuracy is always desirable in face recognition system. Human faces have a complex multidimensional structure so required an efficient and advance method for accurate detection and recognition. It attracts researchers to work in the field of face recognition. Various existing methods e.g. pattern recognition, machine learning, Eigen faces (E.g. PCA), Fisher faces (E.g. LDA) and Laplacian Face (E.g. LAAP). Existing methods encounters with several issues e.g. detection rate, accuracy and time. In this research we are presenting an efficient and more accurate face detection and recognition method based on improved locality preserving projection and semi supervised learning method (ILPP-SSLM). A LPP method is basically an appearance based method. One extra feedback parameters are added in existing LPP which helps to improve the accuracy %. Proposed method firstly finds an embedding to preserve local information and then find a sub face to detect the desired face. Proposed method apply semi supervised learning method which uses combine features of supervised as well as unsupervised learning. Proposed method uses KNN (K-Nearest Neighbor) "supervised" classification method in that it uses the class labels of the training data and LPP dimensionality reduction method (unsupervised learning). An experimental study clearly shows that proposed method performs much better in terms of accuracy %, detection rate % over existing LPP method.