Carnatic Raga Emotion Audio Dataset – RaagaDhvani

The RaagaDhvani dataset comprises high-quality audio recordings of solo flute renditions representing 11 selected Carnatic ragas: Bhairavi, Bilahari, Hamsadhwani, Saveri, Thodi, Anandabhairavi, Poorvikalyani, Gaanamurthe, Neelambari, Bhauli, and Dwijavanthi. The primary content includes audio data along with derived emotion annotations, augmentation variants, and associated metadata. Each raga is intended to capture a specific emotional character, curated to support research in music emotion recognition and affective computing.The dataset is intended for research in music emotion recognition, affective computing, music therapy, and Indian classical musicology.

• Size

The dataset includes over 448 audio samples, generated through original recordings and extensive data augmentation techniques (e.g., pitch shift, time stretch, and noise addition). Each audio sample ranges from 30 to 60 seconds in duration, resulting in a total of approximately 20 to 25 hours of audio data. The dataset ensures a relatively balanced distribution of samples across the 11 ragas and their corresponding emotional classes.

• Format

Audio files: .wav format, recorded in 44.1 kHz, mono, 16-bit resolution to ensure fidelity and uniformity.

Features: MFCC-based audio features are extracted using the librosa library and stored as .npy (NumPy) arrays for efficient use in machine learning workflows.

Metadata and Annotations: Stored in structured formats such as .csv files. These files contain fields like filename, raga, emotion, augmentation_type, and duration.  

All recordings were created by Prof. Vidushi Archana Priyadarshini and are released under the Creative Commons Attribution-NoDerivatives 4.0 International (CC BY-ND 4.0) license.