基于深度卷积神经网络的手写哈萨克文字母识别方法

A Handwritten Kazakh Letter Recognition Method Based on Deep Convolutional Neural Network

利用卷积神经网络在图像识别方面的优势,提出了一种基于深度卷积神经网的哈萨克手写字母识别方法(DCNN-KLR),成功建立了一种哈萨克手写文字识别模型。与传统的方法(SVM+HOG)相比,不仅训练方便、速度快,而且提高了哈萨克手写文字的识别率。在5708个数据样本上进行训练和测试,将样本分为33类和100类,正确识别率分别达到93.29%和92.38%。

This paper proposes a handwritten Kazakh letter recognition method based on deep convolutional neural network（DCNN-KLR）. We exploit the advantages of the convolutional neural network in image recognition to build a model for handwritten Kazakh letter recognition.Compared to traditional method（SVM ＋ HOG）,this method is not only convenient and fast in training but also achieves a higher recognition accuracy. We use 5708 data samples for training and testing. The data is divided into 33 classes with 100 categories,and the highest recognition rate can be up to 93. 29% and 92. 38%.