Putri, Angela Adytha (2026) Penerapan Kombinasi Metode ResNet dan Bi-GRU untuk Klasifikasi Genre Musik Berbasis Fusi Citra Scalogram-Spectrogram. Other thesis, Institut Teknologi Sepuluh Nopember.
![[thumbnail of 5002211131-Undergraduate_Thesis.pdf]](http://repository.its.ac.id/style/images/fileicons/text.png) |
Text
5002211131-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (31MB) | Request a copy |
Abstract
Musik memiliki berbagai macam genre yang berperan penting dalam pengelompokkan dan sistem rekomendasi lagu. Klasifikasi genre musik menjadi salah satu permasalahan dalam bidang pengolahan sinyal audio karena karakteristik sinyal yang kompleks dan bervariasi. Penelitian ini mengusulkan metode klasifikasi genre musik dengan memanfaatkan kombinasi arsitektur Residual Network (ResNet) dan Bidirectional Gated Recurrent Unit (Bi-GRU) menggunakan representasi citra sinyal audio. Sinyal audio pada dataset GTZAN diubah menjadi citra skalogram menggunakan Continuous Wavelet Transform (CWT) dan citra spektrogram menggunakan Short-Time Fourier Transform (STFT). Citra yang dihasilkan selanjutnya dipotong menjadi 10 bagian dan difusikan dengan menggunakan skalogram sebagai lapisan R, spektrogram sebagai lapisan G, serta hasil penggabungan keduanya menggunakan operasi aritmatika dan fungsi agregasi sebagai lapisan B. Citra hasil fusi digunakan sebagai input pada model ResNet-BiGRU. Model ResNet-BiGRU dimodifikasi dengan penambahan lapisan Adaptive Max Pooling, Batch Normalization, dan Dropout setelah proses ekstraksi fitur oleh ResNet. Hasil pengujian menunjukkan bahwa skenario fusi weighted average dengan nilai parameter pembobotan sebesar 0.5 mampu memberikan performa terbaik dengan akurasi validasi sebesar 96.19%.
===================================================================================================================================
Music has various genres that play an important role in music grouping and song recommendation system. Music genre classification is one of the challenges in the field of audio signal processing due to the complex and diverse characteristics of audio signals. This research proposes a music genre classification method by utilizing a combination of Residual Network (ResNet) and Bidirectional Gated Recurrent Unit (Bi-GRU) architectures using audio signal image representations. The audio signals from the GTZAN dataset are transformed into scalogram images using Continuous Wavelet Transform (CWT) and spectrogram images using Short-Time Fourier Transform (STFT). The resulting images are then segmented into 10 parts and fused by assigning the scalogram as the R channel, the spectrogram as the G channel, and the combination of both using arithmetic operations and aggregation functions as the B channel. The fused images are used as an input to the ResNet-BiGRU model. The ResNet-BiGRU model is modified by adding Adaptive Max Pooling, Batch Normalization, and Dropout layers after the feature extraction process performed by ResNet. Experimental results show that the weighted average fusion scenario with a parameter value of 0.5 achieves the best performance, with a validation accuracy of 96.19%.
| Item Type: | Thesis (Other) |
|---|---|
| Uncontrolled Keywords: | Fusi, Genre Musik, Klasifikasi, Skalogram, Spektrogram, Classification, Fusion, Music Genre, Scalogram, Spectrogram |
| Subjects: | M Music and Books on Music > M Music Q Science > QA Mathematics > QA336 Artificial Intelligence Q Science > QA Mathematics > QA403.3 Wavelets (Mathematics) Q Science > QA Mathematics > QA404 Fourier series Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science) Q Science > QA Mathematics > QA76.9.I52 Information visualization Q Science > QA Mathematics > QA9.58 Algorithms T Technology > T Technology (General) > T57.5 Data Processing |
| Divisions: | Faculty of Science and Data Analytics (SCIENTICS) > Mathematics > 44201-(S1) Undergraduate Thesis |
| Depositing User: | Angela Adytha Putri |
| Date Deposited: | 27 Jan 2026 02:01 |
| Last Modified: | 27 Jan 2026 02:01 |
| URI: | http://repository.its.ac.id/id/eprint/130503 |
Actions (login required)
 |
View Item |