Sistem Peringatan Dini Kondisi Kantuk Pengemudi Motor Menggunakan Sensor IMU dan Support Vector Machine Untuk Deteksi Microsleep

Putra Adira, Achmad Bayhaqi (2026) Sistem Peringatan Dini Kondisi Kantuk Pengemudi Motor Menggunakan Sensor IMU dan Support Vector Machine Untuk Deteksi Microsleep. Undergraduate thesis, Politeknik Negeri Jember.

There is a more recent version of this item available.

	Text (Abstract) Abstract.pdf - Submitted Version Available under License Creative Commons Attribution Share Alike. Download (226kB)
	Text (Bab 1 Pendahuluan) Pendahuluan.pdf - Submitted Version Available under License Creative Commons Attribution Share Alike. Download (212kB)
	Text (Daftar Pustaka) Daftar Pustaka.pdf - Submitted Version Available under License Creative Commons Attribution Share Alike. Download (196kB)
	Text (Laporan Lengkap) Skripsi_Final.pdf - Submitted Version Restricted to Registered users only Download (2MB) \| Request a copy

Abstract

Traffic accidents caused by drowsiness and microsleep among motorcyclists represent a significant road safety concern in Indonesia, with 299,479 accidents recorded in the 2023-2024 period, 95% of which were attributed to human error encompassing fatigue, inattention, and microsleep. This study aims to design and implement a real-time drowsiness early warning system integrated into a motorcycle helmet as a wearable device, utilizing an MPU6050 IMU sensor and a Support Vector Machine (SVM) algorithm for microsleep detection. An experimental approach was adopted, encompassing hardware design based on an ESP32 microcontroller embedded in the helmet, head movement data acquisition using a non-overlapping window segmentation method of 10 samples (1 second at 10 Hz sampling frequency), and a four-stage feature selection pipeline comprising statistical significance testing, Cohen's d discriminative power analysis, correlation-based redundancy elimination, and Recursive Feature Elimination with Cross-Validation (RFECV). Five time-domain statistical features were selected as inputs to an SVM model with a Radial Basis Function (RBF) kernel (C=1, gamma=1.0): angleX_max, angleY_min, angleY_mean, gy_min, and gy_max, extracted from a dataset of 4,750 samples. Model evaluation on the test set yielded an accuracy of 92.53% with a microsleep class recall of 92%, while field testing involving 5 subjects across 50 simulated microsleep events produced an overall detection rate of 86% and a mean computational latency of 21.4 ms per window. The developed helmet-based wearable early warning system demonstrates adequate real-time microsleep detection performance for road safety applications, though further generalization across diverse vehicle types and rider physical characteristics necessitates expanded training data collection in future research.

Item Type:

Thesis (Undergraduate)

Contributors: