Digital audio preservation for Indonesian traditional vocal recognition based on machine learning: A literature review and bibliometric analysis

¹ Program Studi Musik, Universitas Pendidikan Indonesia, Bandung, Indonesia
² Program Studi Pendidikan Seni Musik, Universitas Pendidikan Indonesia, Bandung, Indonesia
³ Program Studi Sistem Komputer, Universitas Tanjungpura, Pontianak, Indonesia

^* Corresponding author: dicemidyanti@upi.edu

Abstract

The study aims to save Indonesia’s extensive voice history by comprehensively examining existing literature and doing a bibliometric analysis. This approach provides a comprehensive understanding of this field’s development, methodology, obstacles, and potential future paths. The key focus is machine learning approaches to identify and safeguard Indonesian traditional vocals using several methods, like spectrogram-based techniques, convolutional and recurrent neural networks, transfer learning, attention mechanisms, and hybrid learning. Examining these technologies considers Indonesia’s voice variety, providing insights into their adaptability to handling distinct scales, tunings, and stylistic variances. The study incorporates a bibliometric analysis to measure the expansion of literature and ascertain the prominent authors, journals, and keywords in this developing topic. This study improves our comprehension of the research terrain and the conceptual paths that drive the progress of the field. Indonesia’s traditional vocal music faces the imminent challenges of industrialization and globalization. However, there is hope for developing machine learning to preserve digital audio data of traditional music, especially traditional vocals in Indonesia, some of which are almost extinct. We explore the use of machine learning to honour and protect Indonesia’s varied vocal traditions while also considering the ethical responsibilities associated with this undertaking.