Hotspots for Enhancing Quranic Speech Recognition

Abstract

This paper discusses hotspots where minor changes in these spots can lead to significant changes in results. Speech recognition is a process with multiple stages and in each stage; there are spots to be enhanced. Recording principles and avoiding issues are essential steps. Collecting the relevant data or designing the data in such form or structure can enhance the training model. In addition, extracting phonemes using a smart acoustic model or using a robust language model tool are another important spots for improvement. Finally, using Deep Neural Network (DNN) as a speech recognition model has been proved to achieve very good results.