Neural Network Face recognition,  Sponsored by University of Bahrain,   1997.   (Summary ) 

Summary

People in artificial vision and pattern recognition have been working on automatic recognition of human faces for the last 20 years.  Automatic face recognition has indeed received a high potential of research due to the potential benefits.  Human face recognition has been employed in different commercial and law enforcement applications.   Also it has been employed for mug shots matching,  bank-store security,  crowd surveillance,  expert identification,  witness face reconstruction,  electronics mug shots book,  and electronic lineup.  In an attempt to conduct new research horizon related to face recognition,   a research frame work has established through a funded laboratory setup by the Deanship for the Scientific Research at the University of Bahrain for the year of 97-98.    The research lasted for two and half years,  and now it is ready for testing and use. Hence,  in this  report we present an overview of the major findings of the principal component analysis (PCA) approach to facial analysis,  and presents our research key factor which is the face recognition via PCA.   Faces are represented as a weighted sum of macro-features (eigen-vectors or eigen-faces) extracted from a cross-product matrix of face images.  We show that eigen-vectors representing general categorical information can be estimated using a very small set of faces and that the information they convey is generalizable to new faces.

The approach to the face recognition problem is to combine the Principal Component Analysis (PCA) and Artificial Neural Network.    Faces are vertically oriented frontal view with wide expression change.  They are extracted from images by the face detection technique first.   Face detection technique will eliminate background influence as much as possible.  The face space is described by a set of eigen-faces.    Hence, Neural Networks are used to recognize the face through learning correct classification of these new faces.   At the College of Engineering – University of Bahrain,  a real-time system has been created and tested which combines the face detection and recognition techniques,  hence personal identification.   It continuously takes image from a camera, finds the face inside the image and recognizes it.   A recognition rate of more than 88% has been achieved over real tests.  It is also shown that the face database can be easily expanded to accommodate more individuals.

The research results can be used in a number of potential areas and sectors.    from them : Ministry of  Interior, Ministry of  Health, Banking Sectors ( for security ), Airport and  Seaports,  and Controlled Doors and Departments.