Driver Status Identification from Driving Behavior Signals - Digital Signal Processing for In-Vehicle Systems and Safety - page 51

identification problem for a reduced number of drivers to achieve more realistic
results. The best identification result is obtained as 85.21% with the fusion of gas
and brake pedal pressure signals among the three drivers.
Distraction detection is an important issue because cognitive/stress conditions have
a great influence on driving behavior. We achieve 93.2% of success in detecting driver
behavior signals under no specific task while the random rate is about 52% for ten
drivers. In our database, nearly half of the driving sessions are done under specific task.
Among these tasks, dialog on mobile phone, conversation with passenger on-board,
sign reading, and license plate reading are the most effective ones.
Warning drivers about future incidents is an important application area because
many of the traffic accidents are caused by drivers. In this study, we propose a
method of predicting driving behavior based on gas pedal pressure, brake pedal
pressure, and vehicle velocity signals. Predicting driving behavior signal using past
samples yields encouraging results. We performed driver-dependent and indepen-
dent driving behavior prediction experiments. Although prediction error profiles for
the driver-independent experiment are higher than the driver-dependent experi-
ment, driver-independent driving behavior prediction attains sufficiently low error
rates. Distractive conditions are expected to have a great influence on driving
behavior. Our driving behavior prediction results are also supporting this finding.
Prediction of driving behavior signals under distractive conditions is 20% more
erroneous than prediction under no secondary task.
Acknowledgments This work has been supported by TUBITAK under project EEEAG-104E176
and by the state planning organization of Turkey (DPT) under Drive-Safe project.
References
1. Kurahashi T, Ishibashi M, Akamatsu M (2003) Objective measures to assess workload for car
driving. In: Proceedings of the SICE 2003 Annual Conference, vol 1, pp 270-275
2. Miyajima C, Kusakawa T, Nishino T, Kitaoka N, Itou K, Takeda K (2008) On-going data
collection of driving behavior signals. In: Takeda K, Erdogan H, Hansen JHL (eds) Corpus and
signal processing for driver behavior. Springer Business-Science, New York
3. Wakita T, Ozawa K, Miyajima C, Takeda K (2005) Parametric versus non-parametric models
of driving behavior signals for driver identification. In: Kanade T, Jain AK, Ratha NK (eds)
AVBPA, vol 3546, Lecture Notes in Computer Science. Springer, New York, pp 739-747
4. Miyajima C, Nishiwaki Y, Ozawa K, Wakita T, Itou K, Takeda K, Itakura F (2007) In:
Proceedings of the IEEE, driver modeling based on driving behavior and its evaluation in
driver identification, vol 95, no 2, pp 427-437
5. Miyaji M, Danno M, Oguri K (2008) Analysis of driver behavior based on traffic incidents for
driver monitor systems. In: Proceedings of the IEEE intelligent vehicles symposium,
Eindhoven, pp 930-935
6. Erzin E, Yemez Y, Tekalp AM, Ercil A, Erdogan H, Abut H (2006) Multimodal person
recognition for human-vehicle interaction. IEEE Multimedia 13(2):18-31
7. Ma X, Jansson M (2007) Model estimation for car-following dynamics based on adaptive
filtering approach. In: Proceedings of the IEEE intelligent transportation systems conference
ITSC 2007, Seattle, pp 824-829