Hier finden Sie wissenschaftliche Publikationen aus den Fraunhofer-Instituten.

Capacitive Proximity Sensing Supported Advanced Driver Assistance System

: Frank, Sebastian
: Wang, Xiaofeng; Braun, Andreas

Rüsselsheim, 2014, pp. 158 p.; Appendix 44 p.
Rüsselsheim, Hochschule RheinMain, Master Thesis, 2014
Master Thesis
Fraunhofer IGD ()
Ambient intelligence (AmI); Anthropometrics; Artificial intelligence (AI); Automotive industries; Capacitive sensors; Activity recognition; Business Field: Digital society; Research Area: Human computer interaction (HCI)

This work's general topic is advanced driver assistance systems. In particular, it's about the assisted driver seat adjustment in dependence on anthropometric data, the detection of Out-of-Position postures and the driver drowsiness detection. Already existing systems use sensors like in- and off-cabin cameras to detect drowsiness or require the manual input of anthropometric data to adjust the driver's seat. Contrary to these system's approaches, the aim of this work is to build a system which captures drowsiness symptoms, tracks the head position and captures anthropometric data only by the use of invisible seat integrated capacitive proximity sensors. Still, the aim includes the evaluation of the system's concepts to give direction for further examinations.

The idea is the integration of several capacitive proximity sensors at meaningful positions into a driver's seat. Owing to the fact that these sensors can sense through non-conductive materials, the sensors can be installed invisible under the seat cover. Furthermore, the sensors measure changes in the electric field. Occupants, which are in range of the sensors, change the electric field. Therefore, the sensor values shall give information about the occupant's anthropometry and position. With these anthropometric data, an assisted seat adjustment shall be possible. Especially the movement of the driver's head could give information about the driver's drowsiness.

A first question of this report addresses the driver's anthropometry. What's a proper seat adjustment? Furthermore, what are the symptoms for drowsiness and which could be measured with capacitive proximity sensors? Moreover, what is an Out-of-Position posture? With information about the anthropometrical requirements, the work shows which concepts can meet the demands on the system. Owing to the fact that the system needs evaluation, how shall a prototype be developed with reference to the concepts?

Due to the results of the evaluation, the concepts can satisfy the demands on the system. The ideas which rely on machine learning classifiers result in reliable data. Nevertheless, the different approaches show different demands on the collected data's diversity, which is used to train the algorithms. Besides the machine learning classifiers, many functions of the assisted seat adjustment depend on generic relations between the prototype's sensor system and the occupant's anthropometry. These functions show positive results. Nevertheless, a multiclass SVM approach with discrete adjustment classification could lead to better results, because this approach can include more sensors. Therefore, further obedience between the sensors' data and the anthropometry could be included.

Several functions of advanced driver assistance systems are integrated into the capacitive proximity sensing supported advanced driver assistance system. The evaluation shows that invisibly seat integrated capacitive proximity sensors can sense several symptoms of driver drowsiness. Furthermore, the system can assist the driver's seat adjustment and detect "Out-of-Position" postures. The detection concepts are constrained by several requirements for a proper working system. Consequently, the next step is a further integration of the system into a real car. Supplementary, the evaluation shows that the machine learning concepts require a plenty of miscellaneous data. Hence, a further data collection will improve the systems creditableness. Besides the further data collection and real system integration, the developed prototype can be the basis for a further function development, like the gesture recognition for the control of a multimedia system.