摘要：

The technological progress of the last years regarding the connectivity between cars and between cars and the internet made it possible for modern in-car-infotainment systems to offer a wide range of new features that are formerly known from home entertainment systems or mobile computers like smartphones or tablets. Due to an accompanying increase of the complexity of automotive user interfaces, car manufactures are faced with the challenge to develop new and efcient interaction concepts. Thereby, the brand-specic look-and-feel, the integration of mobile devices as well as the user-specic design - which is imposed by the subsequent and user-triggered installation of applications - need to be considered. In particular, this study focuses on the interaction concept of personalization that deals with tailoring the user interfaces towards the individual needs of the users. Therefore, a novel method for the processing of various sensor data is introduced. It provides future mobile systems e.g. in-car-infotainment systems with the ability to capture the situation-dependent user behavior implicitly, to search for regular situation-dependent user behavior and to use the discovered contextual behavior patterns for the purpose of proactive user interface adaptations. For the rst time, methods for the detection of sequential patterns and well-known techniques of unsupervised clustering are applied within a common process in order to nd similar sequences of human-machine interactions within numerous user-specic and environment-specic sensor data and without the need of an explicit involvement of the user. The service-oriented method is supported by clearly dened rules which determine the way specic human-machine interactions affect the contextual personalization, how captured interactions are grouped to verify their relevance and how situations for the modication of the user interfaces get detected based on the identied contextual preferences. The rule templates and an appropriate prototypical automotive user interface are also introduced as a proof-of-concept along with an extension of the user interface with two different types of contextual personalizations which are adopted in 4 sample automotive use cases. Additionally, a model-based simulation for interaction-specic sensor data is presented. It allows within a lab environment to simulate long term contextual personalizations for the purpose of tests. Finally, this study is completed by a detailed description of the process and the results of an user evaluation. The user evaluation investigates the driver distraction - caused by the contextual personalization - and the overall attractiveness of the adaptive parts of an automotive user interface. It was conducted based on a real-time simulator for contextual sensor data that gets introduced as well.

展开