Yemeni Music Styles Dataset: Audio Recordings, Mel Spectrograms, and Metadata
Description
This dataset contains audio recordings of traditional Yemeni music categorized into five distinct musical styles: Sanaani, Hadrami, Adani, Tihami, and Lahji. Each audio sample is 30 seconds long, stored in WAV format with a sampling rate of 48 kHz and a bitrate of 360 kbps. In addition to the audio files, the dataset includes Mel Spectrogram image representations generated from each audio sample to support deep learning and audio classification tasks. Structured metadata files (CSV format) are provided, containing information such as song title, artist name, genre label, and file path. The dataset was collected and curated manually with the assistance of experts and specialists in Yemeni music as part of a graduation project conducted at the Department of Computer Science and Information Technology, Ibb University, Republic of Yemen. The dataset aims to support research in ethnomusicology, audio signal processing, and artificial intelligence while contributing to the preservation of Yemeni musical heritage.
Files
Steps to reproduce
To reproduce or reuse the dataset, users can download the complete dataset from Mendeley Data. The audio files are organized into folders corresponding to five Yemeni music styles. Each audio sample has a fixed duration of 30 seconds and is stored in WAV format. Mel Spectrogram images were generated from the audio files using standard audio processing techniques, making them suitable for convolutional neural network (CNN)–based classification tasks. Metadata files in CSV format provide labels and descriptive information for each sample. Researchers can directly use the audio files for signal processing or music analysis tasks, the spectrogram images for deep learning experiments, and the metadata files for supervised learning, dataset filtering, or statistical analysis.
Institutions
- Ibb UniversityIbb Governorate, Ibb