شماره ركورد كنفرانس :
4768
عنوان مقاله :
Offline Persian letters recognition based on Histogram of Oriented Gradients features in image
عنوان به زبان ديگر :
Offline Persian letters recognition based on Histogram of Oriented Gradients features in image
پديدآورندگان :
Soleimani Hossein hosseinpbs@gmail.com Department of Electrical and Computer Engineering, Faculty of GhaziTabatabai, Urmia branch, Technical and Vocational University, Urmia, Iran , Ghollamali Alizadeh Alizadeh_88@yahoo.com Department of Electrical and Computer Engineering, Faculty of GhaziTabatabai, Urmia branch, Technical and Vocational University, Urmia, Iran
كليدواژه :
alphabet recognition , feature extraction, feature selection, machine learning
عنوان كنفرانس :
اولين كنفرانس ملي كسب و كارهاي نوين و هوشمند داده كاوي و پردازش تصاوير
چكيده فارسي :
The target of this work is discrimination of Persian alphabet from each other. Hence, in first step, we extract features from images that encode image regions as high dimensional feature vectors that support high accuracy decisions. The required features are extracted from Histograms of Oriented Gradients (HOG). In second step, we use different classifier for learning and testing alphabet classes with partially labeled data which are support vector machine (SVM) with different kernels, K nearest neighborhood (KNN), naïve byes (NB), parabolic neural network (PNN). In the third step, we collect data form 10 Persian people. 5 Persian girl and 5 Persian boys wrote 10 Persian alphabets in paper. After that we created scanning of the handwritten papers to collect alphabet database. We select 70 percent of data for training and 30 percent of data for testing part. Our experimental results show the optimal performance in accuracy, recall, precision and confusion matrix.
چكيده لاتين :
The target of this work is discrimination of Persian alphabet from each other. Hence, in first step, we extract features from images that encode image regions as high dimensional feature vectors that support high accuracy decisions. The required features are extracted from Histograms of Oriented Gradients (HOG). In second step, we use different classifier for learning and testing alphabet classes with partially labeled data which are support vector machine (SVM) with different kernels, K nearest neighborhood (KNN), naïve byes (NB), parabolic neural network (PNN). In the third step, we collect data form 10 Persian people. 5 Persian girl and 5 Persian boys wrote 10 Persian alphabets in paper. After that we created scanning of the handwritten papers to collect alphabet database. We select 70 percent of data for training and 30 percent of data for testing part. Our experimental results show the optimal performance in accuracy, recall, precision and confusion matrix.
