KhalidIbrahem/somali-music-feature-explorer-v2

# Somali Music Feature Explorer # A starter project to extract and visualize audio features from Somali music recordings. It produces waveform, tempo & beats, pitch (PYIN), mel‑spectrogram, and MFCCs, and saves plots and feature arrays. ## Structure ``` somali-music-feature-explorer/ ├── data/ # put short audio clips (e.g., ≤30s for fair use/testing) ├── features/ # saved feature arrays (.npz) ├── reports/ # saved plots (.png) ├── notebooks/ │ └── 01_feature_explorer.ipynb ├── src/ │ ├── extract_features.py │ └── visualize.py └── requirements.txt ``` ## Quickstart 1) Create a virtual env (recommended), then install: ``` pip install -r requirements.txt ``` 2) Put a WAV/MP3 file into `data/` (e.g., `data/song.wav`). 3) Run the notebook `notebooks/01_feature_explorer.ipynb` OR run the CLI: ``` python src/extract_features.py --audio data/song.wav ``` Outputs: - Plots in `reports/`: `waveform_with_beats.png`, `pitch_f0.png`, `mel_spectrogram.png`, `mfcc.png` - Feature arrays in `features/`: `pitch_features.npz`, `mel_spectrogram.npz`, `mfcc_features.npz`, `beat_sync_mfcc.npz` ## Notes - This project uses **librosa** and **matplotlib** only (no seaborn). - Keep audio clips short for quick iteration and fair‑use testing. - Extend this repo with classification or generative models next. # somali-music-feature-explorer-