A Hybridized Feature Extraction Model for Offline Yorùbá Document Recognition

Document recognition is required to convert handwritten and text documents into digital equivalents, making them more easily accessible convenient store. This study combined feature extraction techniques for recognizing Yorùbá in an effort preserve the cultural values heritages of people. Ten were acquired from Kwara State University’s Library, ten indigenous literate writers wrote version documents. These digitized using HP Scanjet300 pre-processed. The pre-processed image served as input Local Binary Pattern, Speeded-Up-Robust-Features Histogram Gradient. extracted vectors Genetic Algorithm. reduced vector was fed Support Vector Machine. A 10-folds cross-validation used train model: LBP-GA, SURF-GA, HOG-GA, LBP-SURF-GA, HOG-SURF-GA, LBP-HOG-GA LBP-HOG-SURF-GA. LBP-HOG-SURF-GA printed gave 90.0% precision, 90.3% accuracy 15.5% FPR. Handwritten document showed 80.9% 82.6% 20.4% (FPR) CEDAR 98.0% 98.4% 2.6% MNIST 99% 99.5% accuracy, 99.0% 1.1% results hybridized extractions (LBP-HOG-SURF) demonstrated that proposed work improves significantly on various classification metrics.