A Deep Learning-Based Offline Speech Recognition System for Real Time Quran Memorization and Correction

Muhammad Mustafa; Usama Sardar; Shams-Ur-Rahman

Authors

Muhammad Mustafa Department of Computer Science, University of Engineering & Technology Mardan, Pakistan
Usama Sardar Department of Computer Science, University of Engineering & Technology Mardan, Pakistan
Shams-Ur-Rahman Department of Computer Science, University of Engineering & Technology Mardan, Pakistan

Keywords:

Quranic Recitation, Automatic Speech Recognition, Offline ASR, Fuzzy String Matching, Arabic NLP

Abstract

For students of Hifz, accurate recitation is non-negotiable—a single misread phoneme can alter the meaning of a verse—yet access to qualified Qari instructors remains scarce across the Muslim world. This paper presents Hifz Master, an offline, real-time speech recognition system that listens to a reciter's voice and flags word-level errors within approximately one second. The system uses the Vosk Arabic ASR engine combined with fuzzy string matching against a verified Quranic text corpus. Evaluation was conducted with 32 participants from 13 cities in Pakistan. Under quiet conditions, Hifz Master achieved a 91% Word Accuracy Rate (WAR) and F1 = 0.90; under ambient noise, performance dropped to 68% WAR and F1 = 0.69. Analysis of 42 labelled phonetic errors identified the ع/غ confusion pair as the dominant error type, accounting for 47.6% of cases (95% CI: 33.0%–62.6%). Inter-rater agreement between two certified Qaris reached 94.3% (κ = 0.88), confirming the reliability of the ground truth labels.

References

“Tarteel AI Quran Memorization.” Accessed: Jun. 16, 2026. [Online]. Available: https://tarteel.ai/

“Quran Apps to Improve Reading, Learning and Memorization.” Accessed: Jun. 16, 2026. [Online]. Available: https://quranacademy.io/apps

“Tajweed Quran Colors Roles - Tajweed Quran.” Accessed: Jun. 16, 2026. [Online]. Available: https://easyquran.com/en/tajweed-quran-colors-roles/

“Challenges in Learning Tajweed and Practical Ways to Overcome Them.” Accessed: Jun. 16, 2026. [Online]. Available: https://najamacademy.com/challenges-in-learning-tajweed-and-practical-ways-to-overcome-them/

Sumayya Al-Fadhli, Hajar Al-Harbi, “Speech Recognition Models for Holy Quran Recitation Based on Modern Approaches and Tajweed Rules: A Comprehensive Overview,” Int. J. Adv. Comput. Sci. Appl., vol. 14, no. 2, 2023, doi: 10.14569/IJACSA.2023.0141297.

“Islam360 - World’s Biggest Quran & Hadith App.” Accessed: Jun. 16, 2026. [Online]. Available: https://theislam360.com/

I. S. Alec Radford, Jong Wook Kim, Tao Xu, Greg Brockman, Christine McLeavey, “Robust Speech Recognition via Large-Scale Weak Supervision,” Open AI, 2024, [Online]. Available: https://cdn.openai.com/papers/whisper.pdf

M. E. Ould, M. Alghamdi, and M. Alkanhal, “Phoneme-based recognizer to assist reading the holy Quran,” Adv. Intell. Syst. Comput., vol. 235, pp. 141–152, 2014, doi: 10.1007/978-3-319-01778-5_15.

“VOSK Offline Speech Recognition API.” Accessed: Jun. 16, 2026. [Online]. Available: https://alphacephei.com/vosk/

“The State of Mobile Internet Connectivity 2025: Overview Report | GSMA Intelligence.” Accessed: Jun. 16, 2026. [Online]. Available: https://www.gsmaintelligence.com/research/the-state-of-mobile-internet-connectivity-2025-overview-report

M. Ali, M. Elshafei, M. Al-Ghamdi, and H. Al-Muhtaseb, “Arabic Phonetic Dictionaries for Speech Recognition,” J. Inf. Technol. Res., vol. 2, no. 4, pp. 67–80, Oct. 2009, doi: 10.4018/JITR.2009062905.

M. Diab and N. Habash, “Natural Language Processing of Arabic and its Dialects,” 2014. Accessed: Jun. 16, 2026. [Online]. Available: https://aclanthology.org/D14-2007/

M. A. Alexei Baevski, Henry Zhou, Abdelrahman Mohamed, “wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations,” arXiv:2006.11477, 2020, doi: https://doi.org/10.48550/arXiv.2006.11477.

Samia Haboussi, Nourredine Oukas, Taha Zerrouki & Halima Djettou, “Arabic speech recognition using neural networks: concepts, literature review and challenges,” J. Umm Al-Qura Univ. Appl. Sci., 2025, [Online]. Available: https://link.springer.com/article/10.1007/s43994-025-00213-w

“A Comparative Study of Pretrained Transformer Models for Quranic ASR: Speech Representations, Label Formats, and Dataset Composition.” Accessed: Jun. 29, 2026. [Online]. Available: https://arxiv.org/html/2606.19747

J. M. Rosana Ardila, Megan Branson, Kelly Davis, Michael Henretty, Michael Kohler, “Common Voice: A Massively-Multilingual Speech Corpus,” arXiv:1912.06670, 2020, [Online]. Available: https://arxiv.org/abs/1912.06670

“Play and Record Sound with Python — python-sounddevice, version 0.5.3.” Accessed: Jun. 16, 2026. [Online]. Available: https://python-sounddevice.readthedocs.io/en/0.5.3/

“PyAudio Documentation — PyAudio 0.2.14 documentation.” Accessed: Jun. 16, 2026. [Online]. Available: https://people.csail.mit.edu/hubert/pyaudio/docs/

“Fuzzy Matching with FuzzyWuzzy: A Comprehensive Guide | by Alpha Iterations | Medium.” Accessed: Jun. 16, 2026. [Online]. Available: https://medium.com/@alphaiterations/fuzzy-matching-with-fuzzywuzzy-a-comprehensive-guide-04873f07de31

“difflib — Helpers for computing deltas — Python 3.14.6 documentation.” Accessed: Jun. 16, 2026. [Online]. Available: https://docs.python.org/3/library/difflib.html

Hanan S. Murayshid, Hafida Benhidour, Said Kerrache, “A Sequence-to-Sequence Approach for Arabic Pronoun Resolution,” arXiv:2305.11529, 2023, [Online]. Available: https://arxiv.org/abs/2305.11529