Musical instrument classification is a crucial aspect of music information retrieval (MIR) and generative music systems. However, the focus has largely been on Western musical traditions, leaving a significant gap in research on non-Western music, particularly Persian music. A recent study aims to bridge this gap by introducing a new dataset of isolated recordings covering seven traditional Persian instruments, as well as two common but originally non-Persian instruments (violin and piano) and vocals.
The researchers propose a culturally informed data augmentation strategy that generates realistic polyphonic mixtures from monophonic samples. This approach is significant because it leverages the unique tonal and temporal characteristics of Persian music, ensuring that the augmented data remains culturally relevant and accurate.
To evaluate their method, the researchers used the MERT model (Music undERstanding with large-scale self-supervised Training) with a classification head. They tested the model on out-of-distribution data, which was obtained by manually labeling segments of traditional songs. The results were promising, with the proposed method yielding the best ROC-AUC (Receiver Operating Characteristic – Area Under Curve) score of 0.795 on real-world polyphonic Persian music.
The success of this method highlights the complementary benefits of tonal and temporal coherence in Persian music recognition. It also underscores the importance of culturally grounded augmentation techniques for robust instrument recognition. This research provides a foundation for more inclusive MIR systems and diverse music generation systems, paving the way for a more comprehensive understanding and appreciation of non-Western musical traditions.
The implications of this research are far-reaching. By improving the accuracy of instrument classification in Persian music, we can enhance music information retrieval systems, making it easier for listeners to discover and enjoy traditional Persian music. Additionally, this research can inform the development of generative music systems that can create diverse and culturally rich musical compositions. Ultimately, this work contributes to a more inclusive and nuanced understanding of the world’s musical traditions.
