P10 - Optimizing the Design of a Multi-Sensor System for On-Line Driver State and Drowsiness Detection
- AHMT 2014 - Symposium des Arbeitskreises der Hochschullehrer für Messtechnik
2014-09-18 - 2014-09-20
- L. Li, K. Thongpull, A. König - Institute of Integrated Sensor Systems, University of Kaiserslautern, Kaiserslautern/D
- 205 - 214
Driver assistance systems have become largely established in automotive applications, both for comfort and for safety functions. The monitoring of the driver state and intention, in particular, the detection of fatigue or drowsiness, is a relevant but not completely satisfying solved task both in consumer as well as commercial vehicles to improve both vehicle and road safety. Thus, in our research, a multi-sensor driver assistance system for this aim has been conceived in our DeCaDrive project . Here, we have advanced the system realization towards higher flexibility, optimization as well as on-line drowsiness detection capability by moving it to the ORANGE multi-platform open-access environment  and employing Support-Vector-Machine classifier  on the DeCaDrive feature data. The employed methods, hierarchical SVM-based classification with automated finding of parameters as well as suitable features, gave comparable or even superior results with regard to previous investigations, as substantially smaller training sets for higher generalization capability were employed. Classification rates of 99.66 % could be achieved for five persons and one hour recorded driving data each.