Pengembangan Aplikasi Klasifikasi Suara Alat Musik Kalimba

Abstract

This study aims to develop an application that implement deep learning with the Convolutional Neural Network (CNN) for classifying the sound of the kalimba and not kalimba. The application development in this research used the incremental method. In increment 1, the dataset will be cut into ten seconds and then converted into a mel-spectrogram image with the help of librosa. The test evaluation results from the experiments carried out were 98.33% accuracy, 0.0394 loss and 98% F1 score with 150 epochs of training. In increment 2, the model is implemented as a GUI with the help of TKinter. This study shows that CNN can be used to classify the sound of the kalimba.