A Novel Multi-Angular LTP and MLDA Based Face Recognition Using Modified Feed Forward Neural Network

Faces represent a complex multidimensional meaningful visual representation which is considered as a credential for an individual's identification and verification. A Feed-Forward Neural Network based face recognition is proposed in order to elevate the effectiveness of the face recognition systems. The convolution neural network provides successively larger features in a hierarchical set of layers. Preprocessing is performed to filter the images prior to feature extraction. The feature extraction is performed via Linear Discriminative Analysis (LDA) and Local Tetra Pattern (LTP) which captures both frequency and location information. Further, feature reduction is performed via (LDA) for image enhancement which enables better image classification. Then, the classification process is performed via M-FNN technique. It estimates the primary component of every class and perceives the classifier based on data projection on subspaces traversed by the principal components. Finally, the performance of the proposed technique is analyzed via parameters such as Accuracy, Sensitivity, Specificity, Precision, and Recall. It is apparent from the results of the performed experiments that the proposed technique offers better results compared to the existing techniques.