Accidents Related to Driver’s Transient Response

The human cognitive states accompanying incorrect/absent motor responses or slow responses in driving tasks on the roads may easily triggers an accident and lead to disastrous consequences as an underlying cause. An impaired driver will not take evasive action prior to a collision, where almost 30% of accidents could be avoided by means of reducing the driver

related reaction time by just 0.5 sec, and thus the reductions in traffic crash losses from reducing crashes attributable to driver impairment far exceed reductions from any other potential countermeasure [53].

1.3.2. Accidents Caused by Driver’s Loss of Alertness

Previous studies have showed that drivers’ fatigue has been implicated as a causal factor in many accidents because of the marked decline in the drivers’ abilities of perception, recognition and vehicle control abilities while sleepy [54-56]. The National Highway Traffic Safety Administration (NHTSA) conservatively estimates that 100,000 police-reported crashes are the direct result of driver fatigue each year [57]. This results in an estimated 1,550 deaths, 71,000 injuries and $12.5 billion in monetary losses. The National Sleep Foundation (NSF) also reported in 2002 [58] that 51% of adult drivers had driven a vehicle while feeling drowsy and 17% had actually fallen asleep. The National Transportation Safety Board found that 58 percent of 107 single-vehicle roadway departure crashes were fatigue-related in 1995, where the truck driver survived and no other vehicle was involved.

Driving under the influences of drowsiness will cause (a) longer reaction time, which will produce effects on crash risk, particularly at high speeds; (b) vigilance reduction including non-responses or delaying responding where performance on attention-demanding tasks declines with drowsiness; (c) deficits in information processing, which will reduce the accuracy and correctness in decision-making tasks.

Therefore, the leading response should be to persuade road users to adopt “error-free”

behavior and maintain the human high performance in the context of road traffic safety

this measure could be further used to predict changes in driver's performance capacity in order to deliver effective feedbacks to maintain their maximum performance.

1.4. Object and Overview of the Thesis

The objects of this thesis are to develop advanced biomedical signal processing methodologies to quantify the level of the human cognitive state with concurrent changes in the driving performance. To achieve these goals, we develop methodologies that combine independent component analysis (ICA), power spectrum analysis, correlation analysis, and fuzzy neural network (FNN) models for ongoing assessment of the transient event-related brain dynamics and the level of alertness of drivers by investigating the neurobiological mechanisms underlying non-invasively recorded multidimensional electroencephalographic (EEG) brain dynamics in the cognitive driving tasks. We then apply these methods to the field of the driving safety, and focus on two major applications, the visual traffic-light detection task and the continuous lane-keeping task on the highway, since they are most frequently happened events on the road in daily life and will easily lead to huge losses in both health injuries and economics.

Three important parts compose this dissertation. The first part describes the research-oriented methodologies for the analysis of human cognitive responses based on EEG signals to support further applications. The second part describes methods for identification of transient brain cognitive responses of drivers related to Red/Yellow/Amber traffic-light events.

The last part focuses on developing methods to monitor alertness level of drivers accompanying changes in driving performances and to explore the relationship between brain activities, and human cognitive states. Each of these parts is described in a separate chapter.

In this chapter, the necessity of developing technologies for driving safety has been described, including statistical reports of road traffic injuries, preventions of traffic accidents

by governments and vehicle-manufactories, and problems of monitoring driver’s cognitive states related to driving errors, which suggested the direction of the thesis.

In Chapter 2, we proposed advanced signal processing methodologies for the analysis of brain activities related to cognitive states, including virtual reality technology, EEG measurement system, independent component analysis, temporal matching filter, moving averaged power spectral analysis, correlation analysis, and fuzzy neural network model.

In Chapter 3, we describe the system architectures focusing on identification of event-related brain potentials related to driver’s transient cognitive responses on traffic-light stimuli, including details of the traffic-light experimental setup, analysis of EEG data using ICA and temporal matching filter, and classification of EEG pattern related to Red/Yellow/Amber stimuli using fuzzy neural networks. Some discussions and conclusion remarks are also included.

In Chapter 4, we propose models for accurately and continuously monitoring level of driver's alertness accompanying changes in driver's performance in a lane-keeping driving task on the virtual-reality-based highway scene. The EEG-based estimation system combines EEG power spectrum, independent component analysis, correlation analysis for adaptive feature selection, and linear regression models and fuzzy neural network estimators. The relationship among EEG power spectrum, driver’s alertness level and driving performance are discussed in detail.

In Chapter 5, contributions of this thesis are summarized with final conclusions. Some further applications using the proposed methodologies are also illustrated.

2. Methodology

In this chapter, a brief overview of the advanced EEG-based biomedical signal processing methodologies for identifying/monitoring driver’s cognitive states is presented. We propose quantitative techniques for ongoing assessment of both the transient event-related brain dynamics and the level of alertness of drivers by investigating the non-invasively recorded EEG brain dynamics in two cognitive driving tasks. Fig. 2-1 shows the whole system architecture consisted of four major parts. (1) The virtual reality technology is used to construct an interactive driving environment for performing two cognitive driving tasks, the visual traffic-light stimulated experiment and the driver’s alertness estimating experiment on highway. (2) The NeuroScan 40-channel EEG measurement system is used to non-invasively collect multidimensional high-fidelity EEG signals. (3) The advanced signal processing technologies are proposed to remove non-brain artifacts, locate optimal positions to wire EEG electrodes, and extract effective features, including independent component analysis, power spectral analysis, correlation analysis, and adaptive feature selecting mechanism. (4) An individual fuzzy neural network model for each subject is used to classify the transient cognitive responses or to monitor the driving performance related to driver’s cognitive states.

ICA/PCA

Figure 2-1. The system architecture of the EEG-based driver’s cognitive-state monitoring system. It consists of four major parts: (1) Virtual-reality-based driving simulator. (2) The NeuroScan EEG measurement system. (3) Advanced signal-processing unit. (4) Fuzzy-neural-network classifier or estimator.