Atiqul Islam RIZVI, Tahmina KHANAM, Kaushik DEB, Ibrahim KHAN, Saki KOWSAR

A comparative study on handwritten Bangla character recognition

Recognition of handwritten Bangla characters has drawn considerable attention recently. The Banglalanguage is rich with characters of various styles such as numerals, basic characters, and compound and modifiercharacters. The inherent variation in individual writing styles, along with the complex, cursive nature of characters,makes the recognition task more challenging. To compare the outcomes of handwritten Bangla character recognition, thisstudy considers two different approaches. The first one is classifier-based, where a hybrid model of the feature extractiontechnique extracts the features and a multiclass support vector machine (SVM) performs the recognition. The secondone is based on a convolution neural network (CNN). For recognition, we considered 10 Bangla numerals, 50 basiccharacters, and a subset of compound characters that are frequently used in the Bangla language. Experimental resultsdemonstrate that the CNN model outperforms the traditional classifier-based approach, obtaining 98.04%, 99.68%, and98.18% recognition accuracy for Bangla basic characters, numerals, and the subset of compound characters, respectively.

PDF

___

[1] Rahman AFR, Rahman R, Fairhurst MC. Recognition of handwritten Bengali characters: a novel multistage approach. Pattern Recogn 2002; 35: 997-1006.
[2] Bhowmik TK, Bhattacharya U, Parui SK. Recognition of Bangla handwritten characters using an MLP classifier based on stroke features. In: 11th International Conference on Neural Information Processing ICONIP; 22–25 November 2004; Kolkata, India. pp. 814-819.
[3] Bhattacharya U, Shridhar M, Parui SK. On recognition of handwritten Bangla characters. In: 5th Indian Conference on Computer Vision, Graphics and Image Processing ICVGIP; 13–16 December 2006; Madurai, India. pp. 817-828.
[4] Bhowmik TK, Ghanty P, Roy A, Parui SK. SVM-based hierarchical architectures for handwritten Bangla character recognition. Int J Doc Anal Recog 2009; 12: 97-108.
[5] Das N, Das B, Sarkar R, Basu S, Kundu M, Nasipuri M. Handwritten Bangla basic and compound character recognition using MLP and SVM classifier. Journal of Computing 2010; 2: 109-115.
[6] Bhattacharya U, Das TK, Datta A, Parui SK, Chaudhuri BB. Recognition of handprinted Bangla numerals using neural network models. In: AFSS International Conference on Fuzzy Systems; 3–6 February 2002; Kolkata, India. pp. 228-235.
[7] Basu S, Sarkar R, Das N, Kundu M, Nasipuri M, Basu DK. Handwritten Bangla digit recognition using classifier combination through DS technique. In: 1st International Conference on Pattern Recognition and Machine Intelligence PReMI; 20–22 December 2005; Kolkata, India. pp. 236-241.
[8] Pal U, Wakabayashi T, Kimura F. Handwritten Bangla compound character recognition using gradient feature. In: 10th International Conference on Information Technology; 17–20 December 2007; Orissa, India. pp. 208-213.
[9] Bag S, Harit G, Bhowmick P. Topological features for recognizing printed and handwritten Bangla characters. In: Joint Workshop on Multilingual OCR and Analytics for Noisy Unstructured Text Data; 17 September 2011; Beijing, China.
[10] Sazal MMR, Biswas SK, Amin MF, Murase K. Bangla handwritten character recognition using deep belief network. In: International Conference on Electrical Information and Communication Technology EICT; 13–15 February 2014; Khulna, Bangladesh. pp. 1-5.
[11] Sharif SMA, Mohammed N, Mansoor N, Momen S. A hybrid deep model with HOG features for Bangla handwritten numeral classification. In: 9th International Conference on Electrical and Computer Engineering; 20–22 December 2016; Dhaka, Bangladesh. pp. 463-466.
[12] Purkaystha B, Datta T, Islam MS. Bengali handwritten character recognition using deep convolutional neural network. In: 20th International Conference of Computer and Information Technology; 22–24 December 2017; Dhaka, Bangladesh. pp. 1-5.
[13] Rabby AKMSA, Haque S, Abujar S, Hossain SA. EkushNet: Using convolutional neural network for Bangla handwritten recognition. Procedia Comput Sci 2018; 143: 603-610.
[14] Lokhande SS, Dawande NA. A survey on document image binarization techniques. In: Proceedings of International Conference on Computing Communication Control and Automation; 2015; Pune, India. pp. 742-746.
[15] Otsu N. A threshold selection method from gray-level histograms. IEEE T Syst Man Cyb 1979; 9: 62-66.
[16] Trier ØD, Jain AK, Taxt T. Feature extraction methods for character recognition-A survey. Pattern Recogn 1996; 29: 641-662.
[17] Bharambe M. Recognition of offline handwritten mathematical expressions. International Journal of Computer Applications 2015; 2015: 35-39
[18] Sinha G, Rani R, Dhir R. Handwritten Gurmukhi numeral recognition using zone-based hybrid feature extraction techniques. International Journal of Computer Applications 2012; 47: 24-29
[19] Dalal N, Triggs B. Histograms of oriented gradients for human detection. In: IEEE Computer Society Conference on Computer Vision and Pattern Recognition CVPR; 20–25 June 2005; San Diego, CA, USA. pp. 886-893.
[20] Li Y, Su G. Simplified histograms of oriented gradient features extraction algorithm for the hardware implementation. In: International Conference on Computers, Communications, and Systems; 2–3 November 2015; Kanyakumari, India. pp. 192-195.
[21] He K, Zhang X, Ren S, Sun J. Deep residual learning for image recognition. In: Proceedings of IEEE Conference on Computer Vision and Pattern Recognition CVPR; 2016; Las Vegas, NV, USA. pp. 770-778.
[22] Biswas M, Islam R, Shom GK, Shopon M, Mohammed N, Momen S, Abedin A. BanglaLekha-Isolated: A multipurpose comprehensive dataset of handwritten Bangla isolated characters. Data in Brief 2017; 12: 103-107.
[23] Sokolova M, Lapalme G. A systematic analysis of performance measures for classification tasks. Inform Process Manag 2009; 45: 427-437.