Analisis Perbandingan Metode Noise Reduction Untuk Deteksi Audio Kendaraan Darurat Menggunakan MobileNetV4 Pada Berbagai Signal To Noise Ratio (SNR)

Pongoh, Fanky Wellsy (2026) Analisis Perbandingan Metode Noise Reduction Untuk Deteksi Audio Kendaraan Darurat Menggunakan MobileNetV4 Pada Berbagai Signal To Noise Ratio (SNR). Bachelor thesis, Institut Teknologi Kalimantan.

[img] Text
11221021_Cover.pdf
Download (157kB)
[img] Text
11221021_statement_of_authenticity.pdf
Download (562kB)
[img] Text
11221021_publishing_agreement.pdf
Download (695kB)
[img] Text
11221021_approval_sheet.pdf
Download (575kB)
[img] Text
11221021_preface.pdf
Download (853kB)
[img] Text
11221021_abstract_id.pdf
Download (262kB)
[img] Text
11221021_abstract_en.pdf
Restricted to Repository staff only until 3 October 2028.
Download (232kB) | Request a copy
[img] Text
11221021_table_of_content.pdf
Restricted to Repository staff only until 3 October 2028.
Download (338kB) | Request a copy
[img] Text
11221021_illustrations.pdf
Restricted to Repository staff only until 3 October 2028.
Download (230kB) | Request a copy
[img] Text
11221021_tables.pdf
Restricted to Repository staff only until 3 October 2028.
Download (227kB) | Request a copy
[img] Text
11221021_notations.pdf
Restricted to Repository staff only until 3 October 2028.
Download (216kB) | Request a copy
[img] Text
11221021_chapter_1.pdf
Restricted to Repository staff only until 3 October 2028.
Download (664kB) | Request a copy
[img] Text
11221021_chapter_2.pdf
Restricted to Repository staff only until 3 October 2028.
Download (2MB) | Request a copy
[img] Text
11221021_chapter_3.pdf
Restricted to Repository staff only until 3 October 2028.
Download (1MB) | Request a copy
[img] Text
11221021_chapter_4.pdf
Restricted to Repository staff only until 3 October 2028.
Download (5MB) | Request a copy
[img] Text
11221021_conclusions.pdf
Restricted to Repository staff only until 3 October 2028.
Download (335kB) | Request a copy
[img] Text
11221021_bibliography.pdf
Download (435kB)
[img] Text
11221021_enclosure.pdf
Restricted to Repository staff only until 3 October 2028.
Download (385kB) | Request a copy
[img] Text
11221021_paper.pdf
Restricted to Repository staff only until 3 October 2028.
Download (723kB) | Request a copy
[img] Text
11221021_presentation.pdf
Restricted to Repository staff only until 3 October 2028.
Download (3MB) | Request a copy

Abstract

Efektivitas peringatan sirine kendaraan darurat sering menurun akibat tingginya kebisingan di lingkungan perkotaan, yang menyebabkan rendahnya Signal-to-Noise Ratio (SNR) dan menyulitkan sistem deteksi audio otomatis. Penelitian ini menganalisis perbandingan efektivitas metode noise reduction (bandpass filter, highpass filter, dan spectral subtraction) terhadap baseline tanpa filtering dalam meningkatkan performa deteksi audio kendaraan darurat menggunakan arsitektur deep learning MobileNetV4. Pengujian menggunakan dataset audio berbentuk mel-spectrogram yang telah diaugmentasi dengan noise injection pada rentang SNR ≤ −30 dB hingga SNR ≥ 30 dB dengan interval 5 dB. Hasil evaluasi menunjukkan bahwa tingkat SNR sangat mempengaruhi performa model, di mana kinerja menurun drastis pada kondisi noise ekstrem (−30 dB dan −25 dB). Bandpass filter mencatatkan kinerja sedikit lebih rendah dibandingkan baseline, sedangkan highpass filter memberikan hasil lebih baik dengan memotong frekuensi rendah secara efektif. Spectral subtraction unggul dan stabil dengan nilai tertinggi pada akurasi, recall, dan F1-Score, khususnya pada SNR rendah (−15 dB dan −10 dB). Namun, seluruh model gagal mengklasifikasi pada noise sangat ekstrem. Kesimpulannya, spectral subtraction cukup efektif mengekstraksi pola sirine pada kondisi normal hingga SNR rendah, tetapi berpotensi gagal pada SNR ekstrem seperti −30 dB.

Item Type: Thesis (Bachelor)
Subjects: Q Science > QA Mathematics > QA75 Electronic computers. Computer science
Divisions: Jurusan Matematika dan Teknologi Informasi > Informatika
Depositing User: Fanky Wellsy Pongoh
Date Deposited: 15 Jul 2026 03:14
Last Modified: 15 Jul 2026 03:14
URI: http://repository.itk.ac.id/id/eprint/27176

Actions (login required)

View Item View Item