Enhancing Education Decision-Making with Deep Learning for Arabic Spoken Digit Recognition

EasyChair Preprint 11572

Abstract

In the realm of education, it becomes imperative to gauge students' learning progress, enabling well-informed decisions and effective support. Recent years have witnessed the ascent of deep learning as a potent instrument for speech recognition. It provides a more exact and efficient examination of people receiving speech treatment. This research offers a deep learning model centered on convolutional neural networks (CNN) suited for the categorization of Arabic spoken digits spanning from 0 to 9. Our model is rigorously trained on a broad dataset including recordings of spoken Arabic numbers, covering authentic Arabic speakers of various ages and ability levels. The findings speak much about the capabilities of our CNN-based algorithm. It attains an outstanding accuracy rate in identifying and classifying Arabic spoken digits, claiming an overall accuracy of 96.10%. Furthermore, we dive into the larger ramifications of our results within the educational environment. This emphasizes the potential of our strategy to better the evaluation of adult learners' speech therapy and to create more effective support measures. This adaptable methodology finds relevance across many educational environments, making voice recognition technology in speech therapy for adult learners more accessible and productive.

Keyphrases: Convolutional Neural Networks, deep learning, speech recognition

@booklet{EasyChair:11572,
  author    = {Zineb Touati-Hamad and Mohamed Ridda Laouar},
  title     = {Enhancing Education Decision-Making with Deep Learning for Arabic Spoken Digit Recognition},
  howpublished = {EasyChair Preprint 11572},
  year      = {EasyChair, 2023}}