Design and analysis of a new way to distinguish faces using neural network technology depending on the Elmans network
Abstract:
Biometrics defined as a measures of the qualities or features unique to humans and is usually used in operations discrimination or electronic identification. And has become a technical basis in many areas because they provide flexibility and high degree of safety in the identification of the true character of the dependence on the personality traits to identify the user, face recognition is one of the Biometric methods which satisfying the high recognition, also has many applications in variety fields, such that confidentiality, image processing, computer vision and video surveillance.
In this work was suggest a system for offline face recognition system, which use Principle Component Analysis (PCA) for (to) select the important features extraction. This paper presents a methodology for offline face recognition system, this method consists of build Elman Neural Network, this proposed system is tested under MATLAB 2010 environment. In general, this work displays the two major cases:
First case (the traditional method): this is the way of the common methods to train neural networks, and through the construction of neural network and one for all people, this system has been tested and gives %86.0 recognition rate.
The second case (the proposed method): This method includes several networks by building a neural network for each person (the number of networks is equal to the persons’ number), This system has been tested and gives %90.0 recognition rate.
Finally, by the comparing between the two cases that the using of one neural network for each person is better than the using of one neural network for all persons.
Keywords: Artificial Neural network; Principle Component Analysis (PCA); Elman Neural Network, feature extraction, face recognition.
الملخص
الكلمات المفتاحية: الشبكات العصبية الاصطناعية, تحليل المركبات الاساسية, شبكة ايلمان العصبية, استخلاص الميزات, تمييز الصور.
الباحثون /
د.شذى عبدالله محمد1
منار عبدالكريم زيدان2
ماهر خلف حسين3
1 قسم هندسة البرمجيات – كلية علوم الحاسوب و الرياضيات – جامعة الموصل – العراق.
2 قسم علوم الحاسوب – كلية التربية للعلوم الصرفة – جامعة الموصل – العراق.
3 قسم علوم الحاسوب – كلية الحدباء الجامعة- العراق
DOI: 10.26389/AJSRP.S130717 عرض البحث كامل عرض العدد كامل