Recognition of compound characters in Kannada language

Abstract

<p><span lang="EN-US">Recognition of degraded printed compound Kannada characters is a challenging research problem. It has been verified experimentally that noise removal is an essential preprocessing step. Proposed are two methods for degraded Kannada character recognition problem. Method 1 is conventionally used histogram of oriented gradients (HOG) feature extraction for character recognition problem. Extracted features are transformed and reduced using principal component analysis (PCA) and classification performed. Various classifiers are experimented with. Simple compound character classification is satisfactory (more than 98% accuracy) with this method. However, the method does not perform well on other two compound types. Method 2 is deep convolutional neural networks (CNN) model for classification. This outperforms HOG features and classification. The highest classification accuracy is found as 98.8% for simple compound character classification. The performance of deep CNN is far better for other two compound types. Deep CNN turns out to better for pooled character classes.</span></p>

Deep convolutional neural network for Lampung character recognition

“…If the two classes are linearly separable, the hyperplane that achieves the maximum margin between them is represented as in ( 9 (11) In this context, C represents the penalty parameter, and its value can enhance classification performance, while 𝛿 𝑖𝑗 denotes the Kronecker symbol. In this investigation, the kernel function utilized is radial based, as in (12):…”

“…Lampung characters originating from documents are shown in Figure 1, while Lampung characters originating from handwriting are shown in Figure 2. Several recognition systems have been proposed for the recognition of document-based and handwritten characters, and a deep learning-based approach has been utilized in many cases, achieving highly promising results in terms of accuracy, Narasimhaiah and Rangarajan [11] conducted an extensive investigation into recognizing Kannada multi-characters. In this study, the dataset underwent classification utilizing the histogram of oriented gradients (HOG) feature descriptor, which was subsequently followed by dimensionality reduction techniques.…”

“…𝑘 }(10)In this context, 𝐾(𝑥 𝑖 , 𝑥 𝑗 ) represents the kernel function, and ∞ 𝑖 signifies the Lagrange multiplier. If the data cannot be linearly separated, the kernel function's mapping is altered as outlined in(11): ISSN: 2302-9285 Bulletin of Electr Eng & Inf, Vol. 13, No.…”