Feature selection based classifier combination approach for handwritten Devanagari numeral recognition

详细信息    查看全文

• 作者：PRATIBHA SINGH ; AJAY VERMA ; NARENDRA S CHAUDHARI
• 关键词：Conditional mutual information maximization (CMIM) ; feature selection (FS) ; minimum redundancy maximum relevance (MRMR) ; mutual information (MI) ; ensemble ; MLP
• 刊名：Sadhana
• 出版年：2015
• 出版时间：September 2015
• 年：2015
• 卷：40
• 期：6
• 页码：1701-1714
• 全文大小：836 KB
• 参考文献：Ahmadzadeh M, Petron M and Sasikala K 2000 The Depster-Shafer combination rule as a tool to classifier combination. Geoscience and Remote Sesing, IEEE International Symposium. Proc. IGARSS, (2429–2431)
  Bagheri M, Montazar G and Kabir E 2013 A subspace approach to error-correcting output coding. Pattern Recog. Lett. 34: 176–184CrossRef
  Bajaj R, Dey L and Chaudhary S 2002 Devnagari numeral recognition by combining decision of multiple connectionist classifiers. Sadhana 27: 59–72CrossRef
  Bhattacharya U and Choudhary B B 2009 Handwritten numeral databases of Indian scripts and multistage recognition of mixed numerals. IEEE Trans. Pattern Anal. Mach. Intell. 31(3): 444–457CrossRef
  Cordella L, De Stefano C, Fontanella F and Marrocco C 2008 A feature selection algorithm for handwritten character recognition. 19th International Conference on Pattern Recognition, (1–4)
  Dash M and Liu H 1997 Feature selection for classification. Intell. Data Anal. 1: 131–156CrossRef
  Dongre V J and Mankar V H 2012 Development of comprehensive devnagari numeral and character database for offline handwritten character recognition. Applied Computational Intelligence and Soft Computing 2012: 1–5CrossRef
  Duch W 2006 Filter Methods, in Feature Extraction: Foundations and Applications. Springer-Verlag New York, Inc. Secaucus, NJ, USA: Studies in Fuzziness & Soft Computing, chapter 3, pp. 89–117
  Fleuret F 2004 Fast binary feature selection with conditional mutual information. J. Mach. Learn. Res. 5: 1531–1555MATH MathSciNet
  Hanmandlu M and Murthy O 2007 Fuzzy model based recognition of handwritten numerals. Pattern Recog. 40: 1840–1854MATH CrossRef
  Kittler J, Hatef M and Duin R W 1998 On combining classifiers. IEEE Trans. Pattern Anal. Mach. Intell. 20 (3): 226–239CrossRef
  Kohavi R and John G H 1997 Wrappers for feature subset selection. Artif. Intell. 97(1–2): 273–324MATH CrossRef
  Kumar R, Kumar A and Ahmed P 2013 A benchmark dataset for devnagari document recognition research. Recent advances in telecommunications, signals and systems, (258–263)
  Kuncheva L I, Bezdek J C and Duin R 2001 Decision template for multiple classifier fusion: An experimental comparision. Pattern Recog. 34: 299–314MATH CrossRef
  Pal U, Wakabayashi T, Sharma N and Kimura F 2007 Handwritten Numeral Recognition of Six Popular Indian Scripts. Ninth International Conference on Document Analysis and Recognition, (749–753) Parana
  Peng H, Long F and Ding C 2005 Feature selection based on mutual information: Criteria of maxdependency, max-relevance, and min-redundancy. IEEE Trans. Pattern Anal. Mach. Intell. 27(8): 1226–1238CrossRef
  Polikar R 2006 Ensemble based systems in decision making. Circuits Syst. Mag. 6(3): 21–45CrossRef
  Shannon C and Weaver W 1949 The mathematical theory of communication. Urbana, IL: University of Illinois Press
  Stefano C D, Fontanella F, Marrocco C and Scotto di Freca A 2014 A GA-based feature selection approach with an application to handwritten character recognition. Pattern Recog. Lett. 35: 130–141CrossRef
  
• 作者单位：PRATIBHA SINGH (1)
  AJAY VERMA (1)
  NARENDRA S CHAUDHARI (2)
  
  1. Department of Electronics and Instrumentation Engineering, Institute of Engineering and Technology DAVV, Khandwa Road, Indore, 452017, India
  2. Computer Science and Engineering Department, Visvesvaraya National Institute of Technology, Nagpur, 440010, India
  
• 刊物类别：Engineering
• 刊物主题：Engineering, general
  
• 出版者：Springer India, in co-publication with Indian Academy of Sciences
• ISSN：0973-7677

文摘

In this paper a method for the recognition of handwritten Hindi numerals is presented. The paper is reporting the effectiveness of the proposed approach, which is utilizing the feature selection based on the Information theory measures. The Multilayer Perceptron (MLP) based classifier combination is used along with feature selection using two criterion functions: (i) Maximum relevance minimum redundancy and (ii) Conditional mutual information maximization. Conditional mutual information based feature selection when driving the ensemble of classifier produces improved recognition results for most of the benchmarking datasets. The improvement is also observed with maximum relevance minimum redundancy based feature selection when used in combination with ensemble of classifiers. The main contribution of the proposed method is that, the method gives quite efficient results utilizing only 10% patterns of the available dataset. Keywords Conditional mutual information maximization (CMIM) feature selection (FS) minimum redundancy maximum relevance (MRMR) mutual information (MI) ensemble MLP