Indian Journal of Science and Technology
Year: 2015, Volume: 8, Issue: 27, Pages: 1-6
Shraddha Arya1*, Indu Chhabra2 and Gurpreet Singh Lehal3
1 Sri Guru Gobind Singh College, Chandigarh - 160019, Punjab, India; [email protected]
2 Department of Computer Science and Application, Panjab University, Chandigarh - 160014, India
3 Computer Science Department, Punjabi University, Patiala - 147002, Punjab, India
Various feature extraction techniques are proposed in literature that can be utilized for recognition purpose in different applications. Gabor Filter is one such well known technique that has the capability to capture image characteristics both in frequency and time domain in parallel. This paper presents an offline Devnagari handwritten numeral recognition system using the famous feature extraction technique of Gabor filter. The effort is to explore the Gabor filter discrimination capabilities and find the optimum feature computation vector in order to improve the recognition rate. Three filter sizes 7 x 7, 19 x 19 and 31 x 31 are experimented with to find the optimal filter size for the given case study. The Standard Benchmark handwritten Devnagari numeral database provided by ISI, Kolkata is used as the training and testing dataset. The original grayscale database images are binarized and normalized to 32 x 32 sizes before feature extraction. The classification is done using the Nearest Neighbor and Support Vector Machine. The maximum recognition accuracy achieved is 98.06%.
Keywords: Devnagari Characters, Gabor Filter, Offline Handwritten Numeral Recognition, SVM
Subscribe now for latest articles and news.