Keywords : back propagation neural network

Online Face Detection and Recognition in a Video Stream

Maha A. Hasso Al-Ghurery; Manar A. Zidan Al-Abaji

AL-Rafidain Journal of Computer Sciences and Mathematics, 2013, Volume 10, Issue 4, Pages 115-134
DOI: 10.33899/csmj.2013.163560

Faces detection and recognition process received considerable attention during the past decade, it is still considered one of the most important studies in the field of Image Processing, Pattern Recognition, and Computer Vision, where it drew the attention of many researchers from both the academic and industrial environments because of the broad scope of its practical applications. Face recognition system is considered as one of the biometric information processing systems that is easy to apply, the necessity of  identifying the identity in the areas of security and surveillance systems made face recognition system one of the most important biometric techniques used in the identification of the individual.
In this work a system is designed for Online Face Detection and Recognition depending on multiple algorithms that are: AdaBoost algorithm for the face detection and the two algorithms Principle Component Analysis (PCA) and Linear Discriminate Analysis (LDA) to extract features and use back propagation neural network in recognition.
This system has been applied on a group of people using different numbers of features extracted from the face. Good recognition ratios have been obtained reached (87.5%)  relying on the key frame in the calculation of the of recognition ratio and (94%) relying on the outcome of the 25 video frames in the calculation of the of recognition ratio.
The System was implemented using graphical interfaces of Visual Studio C# 2010 Language.

Signal Distinction Electroencephalograms (EEG) Using a Back Propagation Neural Network Based On Localized Structural Features Extractions

Najlaa M.I. Safar

AL-Rafidain Journal of Computer Sciences and Mathematics, 2005, Volume 2, Issue 2, Pages 87-100
DOI: 10.33899/csmj.2005.164090

In this research, it localized structural feature selection method has been used as a base of quantifying structural changes with time for Electroencephalograms (EEG) obtained from four states two patient and two healthy with eyes open and eyes closed in both. Then these structural characteristics have been submitted to the back propagation neural network for the purpose of signal distinction  by the intelligent methods. BFGS Quasi-Newton Back propagation   function has been used with the data of the network. It gives good results at testing to the values of features extractions that they have not been training with, and it has been reached to the goal with minimum iteration from other common function that is used with back propagation neural network.
The results for classifying EEG using back propagation neural network show that Alzheimer sick can be detected hardly 100% in many channels in case in taken EEG for the patient with eyes closed. The transformed inputs (from the original data of the signal to the features intentional in the research) are ideally suited for effective classification of EEG data. Recognition rates vary for each EEG channel data between 50-100% correct recognition in the four cases. The follow up method can be useful in several applications including time-series analysis, signal processing and speech recognition.