Designing Musical Instrument Identification Systems Using Lightweight Machine Learning Architectures

Abstract

Musicology is a vast field in which the use of electronic instruments has increased significantly with the development of digital audio workstations and virtual instruments. Selecting the appropriate instrument is crucial for producing high-quality music. The traditional approach to identifying musical instruments relies on manually listening to audio recordings, which is time-consuming and inefficient. Therefore, musical signal processing has become one of the most important tasks in audio signal processing. It enables the identification of instruments in music based on their acoustic features. The proposed work focuses on extracting audio features from WAV files and developing an ensemble model that combines Random Forest and Support Vector Machine classifiers using a voting strategy to identify the instruments used in music. The findings highlight the potential applications of this research in music information retrieval systems and contribute to the growing field of computational musicology by providing a scalable and accurate approach to musical instrument identification, achieving an accuracy of 95%.