Scribd Logo
Upload

Welcome to Scribd!

  • 41 views

    Detection of Driving Fatigue Based On Grip Force On Steering Wheel With Wavelet Transformation and Support Vector Machine

    Uploaded by

    Pirvu
    This document describes a study that aimed to detect driving fatigue based on measuring grip force on a steering wheel. The researchers conducted simulated driving experiments where they collected grip force data from both hands of participants. They used wavelet transformation to extract features from the grip force data and compared the performance of different machine learning classifiers (k-nearest neighbors, linear discriminant analysis, support vector machines) at discriminating between drowsy and awake states based on the grip force data. Their best model was a support vector machine with radial basis function that achieved an average accuracy of 75% at detecting fatigue.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd

    Detection of Driving Fatigue Based On Grip Force On Steering Wheel With Wavelet Transformation and Support Vector Machine

    Uploaded by

    Pirvu
    41 views2 pages
    This document describes a study that aimed to detect driving fatigue based on measuring grip force on a steering wheel. The researchers conducted simulated driving experiments where they collected grip force data from both hands of participants. They used wavelet transformation to extract features from the grip force data and compared the performance of different machine learning classifiers (k-nearest neighbors, linear discriminant analysis, support vector machines) at discriminating between drowsy and awake states based on the grip force data. Their best model was a support vector machine with radial basis function that achieved an average accuracy of 75% at detecting fatigue.

    Original Title

    978-3-642-42051-1_19

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    This document describes a study that aimed to detect driving fatigue based on measuring grip force on a steering wheel. The researchers conducted simulated driving experiments where they collected grip force data from both hands of participants. They used wavelet transformation to extract features from the grip force data and compared the performance of different machine learning classifiers (k-nearest neighbors, linear discriminant analysis, support vector machines) at discriminating between drowsy and awake states based on the grip force data. Their best model was a support vector machine with radial basis function that achieved an average accuracy of 75% at detecting fatigue.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    41 views2 pages

    Detection of Driving Fatigue Based On Grip Force On Steering Wheel With Wavelet Transformation and Support Vector Machine

    Uploaded by

    Pirvu
    This document describes a study that aimed to detect driving fatigue based on measuring grip force on a steering wheel. The researchers conducted simulated driving experiments where they collected grip force data from both hands of participants. They used wavelet transformation to extract features from the grip force data and compared the performance of different machine learning classifiers (k-nearest neighbors, linear discriminant analysis, support vector machines) at discriminating between drowsy and awake states based on the grip force data. Their best model was a support vector machine with radial basis function that achieved an average accuracy of 75% at detecting fatigue.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    of 2

    Detection of Driving Fatigue Based on Grip

    Force on Steering Wheel with Wavelet


    Transformation and Support Vector Machine

    Fan Li1 , Xiao-Wei Wang1 , and Bao-Liang Lu1,2


    1
    Center for Brain-like Computing and Machine Intelligence
    Department of Computer Science and Engineering
    2
    MOE-Microsoft Key Lab. for Intelligent Computing and Intelligent Systems
    Shanghai Jiao Tong University
    800 Dongchuan Road, Shanghai 200240, P.R. China
    [email protected]

    Abstract. This paper proposes an unobtrusive way to detect fatigue


    for drivers through grip forces on steering wheel. Simulated driving ex-
    periments are conducted in a refitted passenger car, during which grip
    forces of both hands are collected. Wavelet transformation is introduced
    to extract fatigue-related features from wavelet coefficients. We compare
    the performance of k-nearest neighbours, linear discriminant analysis,
    and support vector machine (SVM) on the task of discriminating drowsy
    and awake states. SVM with radial basis function reaches the best accu-
    racy, 75% on average. The results show that variation in grip forces on
    steering wheel can be used to effectively detect drivers’ fatigue.

    Keywords: fatigue detection, grip force, wavelet transformation.

    1 Introduction
    The world vehicle population was reported to have surpassed the 1 billion-unit
    mark in 2010 (240 million in U.S. and 78 million in China) [1]. Automobile has
    been becoming the most important necessity for travel. However, accompanied
    with it is that more and more traffic accidents have happened. Driving fatigue
    has long been identified as one of the major causes of traffic accidents. It was
    founded that the crash risk was fourteen time higher for drives who had almost
    fallen behind the wheel [2]. According to National Highway Traffic and Safety
    Administration (NHTSA) report, driver fatigue and drowsiness causes 100,00
    crashes annually, resulting in more than 40,000 injuries. If we can determine the
    onset of driving fatigue, such accidents can be avoided.
    Most of the existing driving fatigue detection methods can be divided into
    three categories [3]:
    1) Physical and physiological data of drivers are used to detect their driving
    fatigue. These include the Electroencephalography (EEG), Electrooculogra-
    phy (EOG) and eye patterns and head movement by video [4]. PERCLOS

    M. Lee et al. (Eds.): ICONIP 2013, Part III, LNCS 8228, pp. 141–148, 2013.
    
    c Springer-Verlag Berlin Heidelberg 2013
    142 F. Li, X.-W. Wang, and B.-L. Lu

    (PERcent eyelid CLOSure), put forward by Carnegie Melon Driving Research


    Center, is one of the most effective measure for driving fatigue detection [5].
    2) Driving performance is indirectly assessed by raw rate, lateral position and
    longitude speed. Daimler Chrysler [6] has developed a detection algorithm
    that jointly analyzed these signals to detect drivers’ fatigue.
    3) Drivers in drowsy status would handle steering wheel and tread pedals (gas,
    brake and clutch) more slowly and improperly than in sober state [3,7,8].
    Drivers would diminish grip force when falling into drowsy, even loosen the
    steering wheel fully, which could easily lead to accident. Thum Chia Chieh [7]
    proposed a statistical method to accumulate the logarithm of probability ratio
    of staying in drowsy or awake state. But the model was too simple to work
    well. Eskandarian and Mortazavi [8] trained an artificial neural network to
    detect drivers’ fatigue, based on the hypothesis that under a drowsy state, the
    steering wheel movements become less precise and larger in amplitude. The
    proposed method had a big default that he did not fully consider time-order
    of steering wheel angle, which would lower detection accuracy.

    Vigilance, the ability to maintain attention and alertness over prolonged periods
    of time, is an effect measurement of fatigue. In this paper, a simulated driving
    system was designed and subjects were asked to finish driving task in a real
    car. We collected drivers’ grip force and response time to an audio signal which
    was used to measure drivers’ vigilance while driving. We proposed an effective
    feature extraction method to extract features from time domain and wavelet
    coefficients through wavelet transformation. We compared the performance of
    three classifiers — support vector machine (SVM), linear discriminant analysis
    (LDA) and k-nearest neighbours (KNN), on the task of discriminating drowsy
    and awake states.

    2 Experiment

    2.1 Platform

    Grip Force Detection


    The setup for grip force detection is shown in Fig. 1. The wheel is fully covered
    by two pieces of force sensors to detect drivers’ force of left and right hands. The
    force sensitive resistor (FSR) was available from Interlink Electronics, which is
    a very thin polymer thick film (PTR) device and will give little influence on
    driving. When force applied to its active surface, its resistance decreases. The
    resistance is converted to voltage through a simple resistance to voltage converter
    circuit. The output voltage Vout characterizes the force exerted on the wheel, and
    is collected by a commercial USB collection card with a multi-channel AD chip
    to PC for analysis. The sampling rate is 100 Hz for each hand’s force.

    Simulated Driving System


    A car has been refitted for experiment. The car’s driving operational devices
    (wheel and pedals) were replaced with Logitech’s simulated controllers, whose

    You might also like