A Hybrid System for Handwritten Character Recognition with High Robustness

In the past few decades, offline recognition of handwritten Indic scripts has received much attention researchers. Although an intensive research been reported for various languages, limited work is carried out Odia, Bangla character owing to their complex shapes and unavailability standard datasets. This paper proposes automated model recognizing both Odia characters numerals, along with numerals maximum optimization efficiency. The proposed primarily deals feature parameters which mainly comprises three parts. Firstly, fast discrete curvelet transform (FDCT) used derive multidirectional features from images. Secondly, PCA LDA reduce dimension vector. are finally subjected least-squares support vector machine (LS-SVM), random forest (RF) classification. effectiveness evaluated over benchmark datasets such as character, numeral numeral. efficacy achieves superiority compared state-of-art techniques. discriminatory prospective FDCT establish more suitable than its counterparts.