Improved Face and Facial Expression Recognition Based on a Novel Local Gradient Neighborhood

Abstract

Computing efficiency is a key in biometric identification systems for automatic facial expression recognition. It was integrated within advanced pattern recognition as an excellent paradigm while users shifted towards underlying patterns. Most existing face recognition models suffer from a low recognition rate and significant execution time. To overcome these drawbacks, we propose a new Local Gradient Neighborhood (LGN) descriptor for effective face and facial expression recognition. Firstly, the LGN components obtained by applying LGN for each block of the face image which is represented by 9-size vector. Secondly, the system concatenates features vectors of different blocks to obtain the final feature vector for the face image. Finally, it applies SVM and KNN techniques to classify the input images. Unlike other similar works, the new proposed descriptor is evaluated on two benchmarks, for face recognition and facial expression recognition respectively. The experimental results show an excellent recognition rate and fast execution time. The recognition rate for the ORL face database is 98.50% and the recognition rate for the JAFEE database is 84.28%. Subject Categories and Descriptors: [I.4.7 Feature Measurement]; [I.5 PATTERN RECOGNITION]: Neural nets General Terms: Local Gradient Neighborhood, Face Expression Recognition, Classification, SVM, Feature Extraction