Digital Signal Processing Reference
In-Depth Information
We suggest that already-proven biometric signals such as speech, fingerprint,
and face recognition should be utilized for driver identification. Although CAN-Bus
signals carry important personal traits, it was found that the performance of a CAN-
Bus-based identification was much lower than other biometric systems (between
83% and 90% recognition accuracy).
The driver-dependent system described here used a GMM/UBM-based structure
for distraction detection. The average distraction detection performance was always
above 70% for all maneuvers. However, the system is not able to recognize neutral
cases better than 70% as well. Therefore, the false alarm rate is expected to be
approximately 30%, which is unacceptable for a final safety application. Again,
after using UTDAT and CCDT tools, better data pools are obtained representing
ground truth in the driving timeline. These tools and a finer analysis on distraction
detection improved the results to 95% for distraction detection as reported in [
9
]
using specific driver performance metrics based on high-frequency content, sample
entropy, and standard deviation.
20.4 Conclusion
CAN-Bus signal analysis performed in long-term time windows open the door to
truly human-centric systems which are capable of recognizing the context/maneu-
ver and detecting distraction which can become an important module in driver
status monitoring and assistance systems. This chapter summarized the recent
findings in CAN-Bus analysis in the UTDrive project during the past 1.5 years.
Two important data mining tools were developed and found to be extremely
beneficial for multimedia data analysis. It was understood that if examined care-
fully, CAN-Bus signals carry important traces on context information and driver
status. This concealed information pieces can be made explicit and interpreted for
the benefit of active safety systems incorporating human factors into system design.
Acknowledgments The authors would like to acknowledge the diligent work in audio/task
transcriptions by CRSS staff Rosarita Lubag.
References
1. CAN-Bus
technical
specifications
from Bosch
http://www.semiconductors.bosch.de/pdf/
2. McRuer D, Weir D (1969) Theory of manual vehicular control. Ergonomics 12:599-633
3. MacAdam C (1981) Application of an optimal preview control for simulation of closed-loop
automobile driving. IEEE Trans Syst Man Cybern SMC-11:393-399
4. Boyraz P, Hansen JHL Active vehicle safety systems based on intelligent CAN-Bus signal
modules, in preparation, to be submitted to IEEE Trans. on ITS, June 09
Search WWH ::
Custom Search