Share
تميز كتابة اليد باستخدام الشبكات العصبية == Handwritten Recognition Using Neural Network
Author name:
يوسف محمد باسل
Supervisor name:
ستار بدر سدخان المالكي | فينوس وزير سماوي
General topic:
Computer Science
Specific topic:
Computer Science
Degree:
Master
University:
Al-Nahrain University - College Of Science - Department Of Computer Science
Language:
English
University location:
Baghdad
Key words:
- Handwritten
- Character recognition
- Neural Network
- Wavelet transformation
- Moments
- Complex moments
- Kohenen Neural Network
- Learning vector quantization
- Recognition rate
First pages:
28T819 - p.pdf
Abstract:
تمييز خط كتابة اليد يعتبر من المواضيع المهمة وذلك بسبب تطبيقاته المهمة اضافة الى قيمته النظرية في مجال تمييز النماذج الصورية ,في هذا البحث استخدمت الشبكات العصبية لتمييز الاحرف المكتوبة , مت م التطبيق لثلاثة انواع : - Kohenen All classes in one network, Kohenen one class in one network, and Learning vector quantization.عملية استخراج الصفات استخدمت التحويل المويجي نوع Haar , كما استخدمت الصفات الهندسية لاستخراج الصفات المميزة للحروف وهي العزم والعزم المعقدة , تم بناء النظام باستخدام لغة فيجوال بيسك 6 , وتم بناء قاعدة بيانات مكونة من 130 نموذج اخذت من 130 شخص .اظهرت النتائج ان التحويل المويجي مع OCON وLVQ قد حققت اعلى معدل تمييز وهو 94% . | The problem of handwritten recognition considered to be very important problem because of its numerous applications and theoretical values in the domain of pattern recognition. In this research, models of Neural Networks are used to recognize written characters, applying Artificial Neural Network (ANN) of three types, which are : - Kohenen All Classes in One Network (ACON), Kohenen One Class in One Network(OCON), and Learning Vector Quantization (LVQ).The feature extraction process made use of Haar Wavelet Transformation to extract the parametric features of the handwritten characters.Also Geometrical features were also used to extract features (Moment and Complex Moment).The system was implemented using Visual Basic Language, database of 130 persons was established, 70 samples from the database were used for training, and the all 130 samples were used for testing the system. The efficiency of the system was tested using the Recognition Rate.The results show that the wavelet transformation with both Kohenen Learning Vector Quantization and Kohenen One Class One Network (OCON) achieves the highest recognition rate in which it scores 94%.
