<![CDATA[Indonesia has a rich and diverse culture, one aspect of which is traditional musical instruments. Bali, as one of the provinces in Indonesia, has unique instruments such as the Gangsa, which is an important part of the Gamelan ensemble. The main challenge in learning traditional musical instruments is the accurate recognition and classification of notes. This study aims to classify Gangsa notes using a Long Short-Term Memory (LSTM) model based on audio feature extraction. Three feature extraction methods were used: Mel-Frequency Cepstral Coefficients (MFCC), Chroma Features, and Mel-Spectrogram. The dataset consisted of 10 tone classes recorded manually from Gangsa bars. The research stages included audio pre-processing, feature extraction, model training, and performance evaluation using accuracy, precision, recall, and confusion matrix metrics. The results show that the MFCC-based model achieved the highest accuracy of 100%, followed by Chroma Features with 98%, and Mel-Spectrogram with 88%. This study shows that the selection of appropriate audio features has a significant effect on tone classification performance. These findings contribute to the application of Artificial Intelligence (AI) in cultural preservation through digital music education.]]>
Copyrights © 2025
