Skip to content

Latest commit

 

History

History
75 lines (63 loc) · 2.92 KB

README.md

File metadata and controls

75 lines (63 loc) · 2.92 KB
Raw

manhuw

manhuw (من هو)
en: 'who is he?'

Manhuw is a Quran Reciter Recognition app. It is a proof of concept for an application capable of recognizing and identifying different Quran reciters from audio recordings. It uses advanced audio processing and machine learning techniques to analyze recitations and match them to known reciters.

Features

  • Voice Recognition: Identify Quran reciters from audio clips.
  • Audio Preprocessing: Includes noise reduction, normalization, and trimming to ensure consistent audio quality.
  • Machine Learning: Utilizes a trained model to distinguish between different reciters' voices.

Installation

# Clone the repository
git clone [repository-url]

# Navigate to the project directory
cd manhuw

# Install required Python packages
pip install -r requirements.txt

Usage

Training

By default, audio training data is included in the audio directory. Example:

audio/
│
├── abdulbasit_abdulsamad/
│   ├── recording1.mp3
│   ├── recording2.mp3
│   └── ...
│
├── ahmed_talib_hameed/
│   ├── recording1.mp3
│   ├── recording2.mp3
│   └── ...

To train the model using these recordings, run:

python3 train.py

This will train the model and output a .joblib model file.
(e.g. 1703241495.179929_trained_model.joblib)

Predicting

python3 predict.py
> Enter model filename: 1703241495.179929_trained_model.joblib
> Loaded model
> Recording for 20 seconds...
> Stop recording...
> Loading recording...
> Preprocessing recording...
> Extracting features...
> Predicting...
> Predicted Reciter: muhammad_al_lohaidan
['muhammad_al_lohaidan']

Known caveats

  • Recording Quality Variance: The model's performance is highly dependent on the quality of the input audio. Recordings made in non-studio environments, especially those with background noise or of lower quality, may not be recognized as accurately.
  • Data Limitations: The training dataset primarily comprises studio-quality recitations. Therefore, the model may not perform as well with live recordings or recitations captured from less controlled environments.
  • Format Differences: While the model has been trained on MP3 format, differences in audio file formats (e.g., WAV vs. MP3) might slightly affect the model's prediction accuracy due to variations in compression and audio quality.
  • Length of Recordings: The model is optimized for shorter audio clips. Longer recordings, especially those with varying styles or multiple reciters, might not be identified as effectively.

Contributing

All contributions are welcomed.

License

GNU General Public License v3.0. Manhuw is provided for the sole purpose of benefiting the Ummah and is intended to be used only for noble, Islamic and ethical purposes. No closed source usage allowed.