Optimization of Hijaiyah Letter Handwriting Recognition Model Based on Deep Learning
Alam Rahmatulloh, Ricky Indra Gunawan, Irfan Darmawan, Randi Rizal, Biki Zulfikri Rahmat · 2022
Hijaiyah handwriting recognition is a challenging research topic. There have been many works and research on character recognition from various languages, but the accuracy value is still being done to improve. Meanwhile, the dataset of handwritten characters with hijaiyah letters is still limited. This study proposes a convolution neural network to recognize and classify hijaiyah writing. The datasets used in this study were Hijja and AHCD. In enhancing the advanced model that has been done previously, we propose the addition of the Adam optimization. In addition, in this study, we have processed both Hijja and AHCD datasets with a composition of 60:20:20. This sophisticated model can improve and be better than the previous model with 91% accuracy results on the Hijja dataset and 98% accuracy on the AHCD dataset. Future work of this work can be made into an application so that the results model that has been built can be used in mobile-based applications.