ITSC 2024 Paper Abstract

Close

Paper ThAT2.5

Zhou, Hongmei (Wuhan University of Technology), Cao, Longtian (Intelligent Transportation Systems Research Center, Wuhan Unive), Tian, Kai (Wuhan University of Technology), Zhang, Hui (Wuhan University of Technology), Huang, Yan (Dongfeng Usharing Technology Company Limited), Huang, Zhen (Wuhan University of Technology), Wang, Xu (Shandong University)

Driver Identification Using Accelerometers Data and Optimized Dataset Construction: A Naturalistic Driving Study

Scheduled for presentation during the Invited Session "Towards Human-Inspired Interactive Autonomous Driving I" (ThAT2), Thursday, September 26, 2024, 11:50−12:10, Salon 5

2024 IEEE 27th International Conference on Intelligent Transportation Systems (ITSC), September 24- 27, 2024, Edmonton, Canada

This information is tentative and subject to change. Compiled on December 26, 2024

Keywords Data Mining and Data Analysis, Human Factors in Intelligent Transportation Systems

Abstract

The development of intelligent connected vehicles provides refined multidimensional data and connected conditions for driving behavior research. Relevant literature suggests that drivers can be distinguished by analyzing their driving behavior disparities. However, existing studies encounter challenges, such as diminished classification accuracy, obstacles in practical implementation, and ambiguity in defining the boundaries of feature extraction methods. This study aims to establish a high-accuracy driver identity identification model with readily accessible data. Driving behavior data, including vehicle speed and three-axis acceleration, were acquired from a field driving experiment involving 20 participants, each driving for a minimum of 6 hours. Driver identification accuracy was assessed by constructing random and continuous datasets and adjusting feature extraction parameters. The K-nearest Neighbor and Convolutional Neural Network models were employed to develop the driver identification model. Evaluate results show that the test accuracy using a random dataset is significantly better than a continuous dataset. When employing the random dataset, the K-Nearest Neighbors algorithm achieved the highest identification accuracy, reaching 93.4%. Notably, within a specific range, reducing the interval of acceleration sampling frequency and incorporating data from multiple axes substantially enhances driver identification accuracy. The research outcomes may facilitate the practical application of the driver identification model.

 

 

All Content © PaperCept, Inc.


This site is protected by copyright and trademark laws under US and International law.
All rights reserved. © 2002-2024 PaperCept, Inc.
Page generated 2024-12-26  05:57:12 PST  Terms of use