YORÙBÁ CHARACTER RECOGNITION SYSTEM USING CONVOLUTIONAL RECURRENT NEURAL NETWORK

Handwritten recognition systems enable automatic of human handwritings, thereby increasing human-computer interaction. Despite enormous efforts in handwritten recognition, little progress has been made due to the variability handwriting, which presents numerous difficulties for machines recognize. It was discovered that while tremendous English and Arabic languages, very work done on Yorùbá characters. Those few works, turn, use Hidden Markov Model (HMM), Support Vector Machine (SVM), Bayes theorem, decision tree algorithms. To integrate save one Nigeria's indigenous languages from extinction, as well make documents accessible available digital world, this research undertaken. The a convolutional recurrent neural network (CRNN) Data were collected students Kwara State University who literate writers. data subjected some level preprocessing such grayscale, binarization, normalization order remove perturbations introduced during digitization process. model trained using preprocessed images. evaluation conducted acquired characters, 87.5% images used training 12.5% evaluate developed system. As there is currently no publicly database characters validating systems. resulting accuracy 87.2% with under dot diacritic signs low accuracy.