A PERCLOS-based Driver Fatigue Recognition
Application for Smart Vehicle Space
Wu Qing, Sun BingXi, Xie Bin and Zhao Junjie
College of Computer Science, Hangzhou Dianzi University
Hangzhou Zhejiang, P.R. China, 310018
wuqing@hdu.edu.cn
Abstract—The paper selected PERCLOS to evaluate
driving fatigue after the comparison of various fatigue
detection methods for smart vehicle space. We detected
driver fatigue status by measuring the proportion of
eyes closed in a certain period of time and the continued
closure time. On the basis of the Haar-Like feature,
AdaBoost algorithm was adopted to produce the strong
classifier for face and eye detection. AdaBoost detector
is employed firstly to determine human face region,
locate the eyes in this region, and using an improved
template matching method to detect eye States.
Experiments show that this method can identify eye
state rapidly and real-timely under natural light
conditions, the algorithm has better robustness and
real-time.
Keywords—Smart Vechilce Space; Driver fatigue; PERCLOS
I. INTRODUCTION
With the research and development of Smart Vehicle, all
kinds of vehicle sensor (microwave radar, laser radar, camera,
other type sensors and so on) and computer equipment are used
in Smart Vehicle Space. We obtained the perception of
vehicles ˈ driving environment and driver state by using
various sensors. At presentˈthe study of high-performance
intelligent perception and pre-warning technology of driving
fatigue has become the hotspot [1]. Driver fatigue monitoring
method can be broadly divided into physiological signals of
drivers, physiological response characteristics of drivers, driver
behavior and vehicle operating status information [2].The test
based on the driver's physiological signals is to test the driver's
brain signals, such as EEG and ECG, This method of fatigue
recognition
has high accuracy , but physiological signals need
to use contact measurement, and more reliance on individual, It
has many limitations for the driver fatigue monitor, so it is
mainly used in the experimental stage, as the experimental
control parameters. The method based on driver behavior and
vehicle status detection used operating characteristics of the
driver's steering wheel and track changes to speculate fatigue
of the driver, but it is subject to personal habits, speed, road
conditions, operating skills, state of the vehicle, roads, and
many other environmental factors. Physiological response
characteristics is use of driver's eye movement and head
movements characteristics to infer the driver's fatigue state,
Commonly used non-contact measurement, It is usually using
the camera to obtain information on the driver's face, through
digital image processing for real-time recognition of eye state
and in turn determine the driver's fatigue state[3].Accuracy and
usefulness of this identification method is both good, Drivers
detection method based on physiological response
characteristics is widely used now.
II. S
MART VEHICLE SPACE ARCHITECTURE
In smart space, the information space can interact with
physical world naturally and spontaneously. In order to get
better communication and collaboration with each other, smart
space entities should take some self-adjusting actions to change
the composition of structure and the function according to the
user, environment and equipment status. The Smart vehicle
space is a computing platform, which is built on the top of
integrated hardware and software technology, based on a
common automotive physical environment. Here we introduce
the smart vehicle space platform from two aspects of the
hardware architecture and software framework.
A. The Hardware Platform Architecture
The computing environments in smart vehicle space
contain a great number of computing devices. The smart
vehicle space hardware includes the following four parts:
Central control unit, which is the core of the hardware manager,
is responsible for the control and command the other parts;
Electronic control unit, which is used to collect and transmit
the data from the sensors, receives and delivers the command
in the smart vehicle space network; intelligent sensors, a kind
of tiny intelligence equipments used to collect the context
information. Intelligent sensors, different with ordinary sensors,
have a self-management and self-communication capability.
The hardware platform consists of the vehicle control
center, data collection devices, remote data servers, the output
devices and automobile bus data acquisition facilities. Data
collection devices, including: PDA, for directly interact with
user, which represents all of the user's hand-held devices such
as mobile phones, personal digital assistant devices and other
portable intelligent devices that can record the user's personal
information and schedule; Cameras, installing more than one
camera in the vehicle space, the general USB cameras can
collect the user's facial features to identify the driver's identity
and detect the fatigue state; the body cameras, around the
vehicle space, allow you to get the vehicle surrounding area
information, not only can solve the problem of the rear view
blind spot, but also monitor the state of vehicle body;
Third International Symposium on Information Processing
978-0-7695-4261-4/10 $26.00 © 2010 IEEE
DOI
437