A mobile learning prototype rates Quran reading practice from 0 to 100 instead of true or false.
Learning to read Hijaiyah letters, the first step toward reading the Quran, usually depends on a mentor. The Iqro method works well, but it needs an Ustaz or Ustazah sitting beside the learner to judge whether a letter was pronounced correctly. Time and availability make that hard to arrange every day.
Most Hijaiyah apps on the market lean on vocabulary, writing, and reading drills. When they do score pronunciation, they hand the audio to a general speech recogniser and return a simple correct or incorrect verdict. The authors argue that this hides progress, because a learner cannot see whether they are getting closer.
The prototype extracts pitch, volume, and rhythm from the learner's voice and folds them into a single score between 0 and 100. The approach borrows from automatic singing evaluation, where pitch contours are aligned with dynamic time warping before rhythm is measured. A field study with five early participants shaped the design, and two requests stood out: a real number instead of a verdict, and a flow where the learner simply speaks to the app.
Ten respondents read a set of letters aloud. Each attempt was scored twice, once by an Ustaz and once by the application. Comparing the two sets of results gave a Pearson correlation coefficient of 0.51 for Iqro level 1 material.
That is a moderate agreement, not a match. In the table the app consistently scored higher than the human judge, which suggests the feature weights still need tuning. The authors describe the work as a feasibility study and plan to test timbre as a fourth feature, collect more data, and extend the method to higher Iqro levels.
Komang Candra Brata, Mohamad Handy Nugraha, and Adam Hendra Brata (2022). An alternative m-learning scoring approach to automatically evaluate Iqro's pronunciation based on pitch, volume, and rhythm features. International Journal of Electrical and Computer Engineering.
