Analisis Robustness Model GraFPrint Terhadap Manipulasi Pitch Shifting dan Time Stretching pada Sistem Identifikasi Hak Cipta Lagu - Submit Jurnal

Fakhrurrozi, Ahmad (2026) Analisis Robustness Model GraFPrint Terhadap Manipulasi Pitch Shifting dan Time Stretching pada Sistem Identifikasi Hak Cipta Lagu - Submit Jurnal. Bachelor thesis, Institut Teknologi Kalimantan.

	Text 11221011_cover.pdf Restricted to Repository staff only until 3 October 2028. Download (314kB) | Request a copy
	Text 11221011_statement_of_authenticity.pdf Restricted to Repository staff only until 3 October 2028. Download (70kB) | Request a copy
	Text 11221011_publishing_agreement.pdf Restricted to Repository staff only until 3 October 2028. Download (187kB) | Request a copy
	Text 11221011_approval_sheet.pdf Restricted to Repository staff only until 3 October 2028. Download (47kB) | Request a copy
	Text 11221011_preface.pdf Restricted to Repository staff only until 3 October 2028. Download (226kB) | Request a copy
	Text 11221011_abstract_id.pdf Restricted to Repository staff only until 3 October 2028. Download (241kB) | Request a copy
	Text 11221011_abstract_en.pdf Restricted to Repository staff only until 3 October 2028. Download (240kB) | Request a copy
	Text 11221011_table_of_content.pdf Restricted to Repository staff only until 3 October 2028. Download (251kB) | Request a copy
	Text 11221011_illustrations.pdf Restricted to Repository staff only until 3 October 2028. Download (239kB) | Request a copy
	Text 11221011_tables.pdf Restricted to Repository staff only until 3 October 2028. Download (228kB) | Request a copy
	Text 11221011_notations.pdf Restricted to Repository staff only until 3 October 2028. Download (267kB) | Request a copy
	Text 11221011_chapter_1.pdf Restricted to Repository staff only until 3 October 2028. Download (324kB) | Request a copy
	Text 11221011_chapter_2.pdf Restricted to Repository staff only until 3 October 2028. Download (1MB) | Request a copy
	Text 11221011_chapter_3.pdf Restricted to Repository staff only until 3 October 2028. Download (553kB) | Request a copy
	Text 11221011_chapter_4.pdf Restricted to Repository staff only until 3 October 2028. Download (1MB) | Request a copy
	Text 11221011_conclusions.pdf Restricted to Repository staff only until 3 October 2028. Download (258kB) | Request a copy
	Text 11221011_bibliography.pdf Restricted to Repository staff only until 3 October 2028. Download (250kB) | Request a copy
	Text 11221011_enclosure.pdf Restricted to Repository staff only until 3 October 2028. Download (504kB) | Request a copy
	Text 11221011_paper.pdf Restricted to Repository staff only until 3 October 2028. Download (1MB) | Request a copy
	Text 11221011_presentation.pdf Restricted to Repository staff only until 3 October 2028. Download (3MB) | Request a copy
	Text 11221011_Form. TA-020.pdf Restricted to Repository staff only until 3 October 2028. Download (105kB) | Request a copy

Abstract

Pesatnya perkembangan industri musik digital memicu maraknya manipulasi audio kreatif yang berpotensi melanggar hak cipta, seperti pitch shifting dan time stretching. Metode audio fingerprinting konvensional berbasis puncak spektral sering kali gagal mengidentifikasi audio yang dimanipulasi secara non-linear akibat hilangnya sinkronisasi koordinat frekuensi statis. Penelitian ini bertujuan mengevaluasi efektivitas model GraFPrint berbasis Graph Neural Networks (GNN) dalam meningkatkan ketahanan sistem identifikasi audio terhadap manipulasi tersebut. GraFPrint memodelkan hubungan struktural antar komponen audio melalui konstruksi graf k nearest neighbor (k-NN) yang dilatih menggunakan self supervised contrastive learning. Penelitian menggunakan dataset Free Music Archive (FMA) dengan pemisahan data pelatihan menggunakan subset FMA small dan pengujian menggunakan subset FMA medium. Evaluasi dilakukan melalui tiga skenario, yaitu kueri pendek berdurasi 2,5 detik, kueri 10 detik menggunakan model pre-trained, dan model hasil fine-tuning. Pengujian dilakukan terhadap manipulasi pitch shifting pada rentang −2 hingga +2 semitone dan time stretching pada rasio 0,8 kali hingga 1,2 kali. Kinerja sistem diukur menggunakan metrik Cosine Similarity, Precision, Recall, dan F1-Score. Hasil penelitian menunjukkan bahwa GraFPrint memiliki ketahanan temporal yang sangat tinggi terhadap manipulasi time stretching, dengan rata-rata Cosine Similarity di atas 0,96 dan F1-Score mencapai 1,0000 pada skenario kueri 10 detik. Namun, performa terhadap pitch shifting mengalami penurunan akibat perubahan struktur spasial pada representasi Log-Mel Spectrogram, dengan Cosine Similarity turun hingga kisaran 0,47–0,63 pada model pre-trained. Proses fine-tuning berhasil meningkatkan ketahanan terhadap pitch shifting dengan F1-Score mencapai 0,9963 dan Cosine Similarity yang membaik secara signifikan, sekaligus mempertahankan F1-Score sempurna pada time stretching. Penelitian ini membuktikan bahwa arsitektur GraFPrint efektif digunakan sebagai dasar sistem manajemen hak cipta musik digital yang adaptif terhadap manipulasi audio modern.

Item Type:	Thesis (Bachelor)
Subjects:	Q Science > QA Mathematics > QA75 Electronic computers. Computer science
Divisions:	Jurusan Matematika dan Teknologi Informasi > Informatika
Depositing User:	Ahmad Fakhrurrozi
Date Deposited:	15 Jul 2026 05:37
Last Modified:	15 Jul 2026 05:37
URI:	http://repository.itk.ac.id/id/eprint/26460

Actions (login required)

View Item