• Betselot Yewulu Reta
• Dhara Rana
• Gayatri Viral Bhalerao
• Melkye Wereta Tsigie

Handwritten character recognition is one of the most challenging problem in the area of pattern recognition. Since different persons have different writing styles. And also, Amharic characters are large in number and some of the characters shape are similar with minor change. In this paper, the basic character recognition techniques performed, such as preprocessing, feature extraction and classification. After preprocessing, haar wavelet transform followed by histogram-oriented gradients feature extracted in each character image. Using haar wavelet transform the character image decomposed into four coefficients such as average, horizontal, vertical and diagonal. Then, HOG feature extracted from four coefficients. The final feature vector created as a single feature vector by concatenating the HOG features extracted from each coefficient. The features used by HOG are gradient orientation and magnitude. Feature extraction method is the most important step. Since classification algorithm depends on the extracted feature of the character image. The dimension of feature is reduced using LDA. We adopted a special type of multi-class SVM in ECOC framework with one versus all design coding matrix. The algorithm is trained and tested on Amharic handwritten characters data set and Chars74K benchmark numeric data set. The proposed model is validated using 10-fold cross-validation technique. As a result, Multi-class SVM classification algorithm and haar wavelet followed by HOG feature extraction technique have been achieved good result in recognizing Amharic handwritten characters.

Amharic handwritten character recognition; Error Correcting Output Code; Haar Wavelet transform; histogram of oriented gradients; liner discriminate analysis; optical character recognition; support vector machine;

"Amharic Handwritten Character Recognition using Machine Learning Approach ", International Journal of Emerging Technologies and Innovative Research (www.jetir.org), ISSN:2349-5162, Vol.5, Issue 5, page no.659-670, May-2018, Available :http://www.jetir.org/papers/JETIR1805300.pdf

"Amharic Handwritten Character Recognition using Machine Learning Approach ", International Journal of Emerging Technologies and Innovative Research (www.jetir.org | UGC and issn Approved), ISSN:2349-5162, Vol.5, Issue 5, page no. pp659-670, May-2018, Available at : http://www.jetir.org/papers/JETIR1805300.pdf