Klasifikasi Sub-Genre Musik Dangdut Menggunakan Jaringan Saraf Tiruan Long Short-Term Memory

Abstract

The automatic classification of Dangdut music sub-genres (Klasik, Koplo, and Campursari) presents a significant challenge in the field of Music Information Retrieval (MIR) due to their overlapping yet distinct musical characteristics. This research proposes a classification system based on a Recurrent Neural Network (RNN) with a Long Short-Term Memory (LSTM) architecture to address this problem. The model is trained using Mel-Frequency Cepstral Coefficients (MFCC) audio features to represent the spectral and timbral characteristics of each sub-genre. The LSTM architecture was chosen for its superior ability to learn temporal dependencies from the sequence of MFCC features. By modeling the evolution of timbre over time, the system can recognize the distinctive patterns that differentiate between Dangdut Klasik, Koplo, and Campursari. The proposed system aims to provide an accurate and efficient classification method, contributing to practical applications such as music recommendation and digital archiving.