Pengembangan Sistem Pendeteksi Kebohongan Berbasis Low-cost Wearable Photoplethysmography Sensor dengan Support Vector Machine

Mizoguchi, Kentaro Mas'ud (2025) Pengembangan Sistem Pendeteksi Kebohongan Berbasis Low-cost Wearable Photoplethysmography Sensor dengan Support Vector Machine. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5026211007-Undergraduate_Thesis.pdf Restricted to Repository staff only Download (4MB) | Request a copy

Abstract

Penelitian ini mengembangkan sistem deteksi kebohongan non-invasif berbasis sinyal Photoplethysmography (PPG) wearable dan klasifikasi Support Vector Machine (SVM). Eksperimen dilaksanakan dengan empat sesi pertanyaan dengan pengambilan sinyal PPG menggunakan sensor Polar Verity Sense. Data diproses melalui pemangkasan data, filtrasi Butterworth, normalisasi data, deteksi puncak, denoising data, kemudian ekstraksi fitur data. Fitur HRV domain waktu (NN, SDNN, SDSD, RMSSD, pNN20/pNN50) dan frekuensi (LF, HF, LF/HF, LF normal, HF normal, LnHF, puncak LF/HF) diekstraksi dari setiap subjek yang kemudian akan dilabelkan untuk klasifikasi model. Model SVM dilatih dengan Leave One Out Cross Validation dan dioptimalkan dengan Grid Search dan Top-K Selection. Hasil optimalisasi memberikan 6 fitur yang digunakan oleh model SVM (NN count, pnn20, HF power, LF/HF, LF normal, HF normal) dengan akurasi terbaik bernilai 72,05%. Disimpulkan bahwa model SVM RBF menunjukkan keseimbangan optimal antara sensitivitas parasimpatis dan simpatis dalam implementasi deteksi kebohongan.
=====================================================================================================================================
This study develops a non-invasive lie detection system based on wearable Photoplethysmography (PPG) signals and classification using Support Vector Machine (SVM). The experiment was conducted across four sessions, during which PPG signals were acquired using the Polar Verity Sense sensor. Data processing included trimming, Butterworth filtering,data normalization, peak detection, data denoising, and data feature extraction. Time-domain HRV features (NN, SDNN, SDSD, RMSSD, pNN20/pNN50) and frequency-domain features (LF, HF, LF/HF, normalized LF, normalized HF, LnHF, LF/HF peak) were extracted for each subject and subsequently labeled for classification. The SVM model was train using Leave One Out Cross Validation and optimized via Grid Search and Top-K Feature Selection. Optimization results six features used by the SVM model (NN count, pNN20, HF power, LF/HF, normalized LF, normalized HF), achieving the highest accuracy of 72.05%. It is concluded that the SVM model with an RBF kernel demonstrates an optimal balance between parasympathetic and sympathetic sensitivity in the implementation of lie detection

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Deteksi Kebohongan, Support Vector Machine, Pembelajaran Mesin, Fotopletismografi, Variabilitas Detak Jantung, Lie Detection, Support Vector Machine, Machine Learning, Heart Rate Variability, Photoplethysmography
Subjects:	T Technology > T Technology (General) > T57.8 Nonlinear programming. Support vector machine. Wavelets. Hidden Markov models.
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Information System > 57201-(S1) Undergraduate Thesis
Depositing User:	Kentaro Mas'ud Mizoguchi
Date Deposited:	25 Jul 2025 01:37
Last Modified:	25 Jul 2025 01:37
URI:	http://repository.its.ac.id/id/eprint/121543

Actions (login required)

View Item