Klasifikasi Instrumen Musik Menggunakan Metode Machine Learning

Abstract

Music plays an important role in human life, and automatic identification of musical instruments is becoming an increasingly relevant field in the digital era. This study aims to classify musical instrument types based on acoustic features using machine learning methods, specifically Support Vector Machine (SVM). The dataset used contains audio recordings of four instruments, namely guitar, piano, drum, and violin. Each audio file goes through a preprocessing process such as sample rate standardization, duration trimming, and framing. Furthermore, feature extraction is carried out from the time domain (Zero Crossing Rate and RMS), frequency domain (Spectral Centroid, Spread, and Roll-off), and cepstral domain (MFCC). The SVM model is trained with a combination of various features and evaluated using accuracy, precision, recall, and F1-score metrics. The experimental results show that the combination of all features produces the best accuracy of 68.33%. Although its performance is not optimal, these results show the potential of a feature-based approach for musical instrument classification and become the basis for further development using more complex methods such as deep learning.