<![CDATA[Wind instruments such as saxophone, clarinet, trumpet, and others possess unique characteristics in terms of frequency, amplitude, and waveform, serving as distinct acoustic signatures for each instrument. In the digital era, the classification of musical instruments for composition and arrangement is often still carried out manually, which can lead to inefficiencies and potential errors in the music production workflow. Automating this process can significantly enhance the speed, accuracy, and consistency of instrument sound classification, especially for large-scale or real-time applications. This study aims to develop a classification system for wind instrument sounds using the Recurrent Neural Network (RNN) method, leveraging acoustic features such as Mel-Frequency Cepstral Coefficients (MFCC) and spectrograms. The RNN architectures employed in this research are Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU), chosen for their ability to capture temporal dependencies in sequential data such as audio signals, making them well-suited for sound-based classification tasks. The dataset consists of 1,200 audio files (.wav) comprising four wind instrument classes: trumpet, baritone, mellophone, and tuba, with 300 samples each. Experimental results show that the LSTM model achieved an accuracy of 95%, while the GRU model reached 99%. Compared to previous studies that reported lower accuracy or focused on broader instrument categories, this research demonstrates a significant improvement in wind instrument classification. The results highlight the effectiveness of RNN-based models in learning the temporal dynamics of audio signals, offering a reliable solution for automated instrument classification in digital music systems.]]>
Copyrights © 2025
