Options
2021
Doctoral Thesis
Titel
Analysis of Capacitive Proximity Sensing as Basis for Human Vehicle Interfaces
Abstract
People spend a lot of time in vehicles. Driving involves risks that are mitigated by passive and active safety systems. Active safety systems prevent accidents from happening in the first place. Human machine interfaces (HMI) are needed to monitor the driver's behavior and capture driver's input. This presents challenges based on environmental, vehicle interior, and user characteristics. Vehicles are directly exposed to the environment. Sensors have to cope with both rapid light changes and complete darkness. Vehicle interiors contain geometries that obscure parts of the human body. Sensors that require a line of sight may therefore be at a disadvantage. Sensor positioning is also limited by vehicle geometry. Sensors that require additional mounting impact the design and result in an obtrusive system. Systems that can be integrated invisibly into existing vehicle structures are less obtrusive and have less impact on the design. Parts of the user must also be monitored in free air, which makes contact-based scanning systems unsuitable. Some sensor systems are capable of monitoring the entire body, but still have to deal with the requirements of vehicle users. People will wear different clothing or glasses. User monitoring must be enabled regardless of this condition. People are also becoming sensitive to their personal data. This can be crucial for the acceptance of systems. Systems must also comply with regulations such as "Privacy by Design" which is required in the European Union. Privacy must therefore be preserved. I argue that capacitive proximity sensors are capable of dealing with the above challenges. Due to the physical principle, illumination changes are not an issue, and they can sense through insulators. Capacitive proximity sensors (CAPS) can therefore be used in existing vehicle structures, both in close proximity and without contact with the object to be monitored. In addition, they are often said to maintain privacy. Based on the challenges and capabilities of CAPS, three research questions emerge: BL RQ1: How can we use existing vehicle structures to enhance or substitute vehicular HMI using CAPS? BL RQ2: How can we use existing vehicle structures to provide new ways of human computer interaction using CAPS? BL RQ3: Can CAPS contribute to the acceptance of vehicular HMI with regard to privacy concerns? To find evidence to support these research questions, I focused on systems that help users drive safely. Cameras are commonly used in HMI. Because they require line of sight, affect interior design, and capture data that creates privacy concerns, they may not be the best choice. CAPS are therefore an opportunity to change the modality. Several applications are developed that provide evidence that the use of CAPS is beneficial for vehicle HMI. Each application is developed following a common process with the goal of meaningful uses. Each application is based on accident statistics and related research, so that real-world problems are addressed. This entails analysis of driving issues, prototype implementations of sensor topologies, and algorithms for attention monitoring, child monitoring, authentication, and gesture recognition in vehicles. One will additionally receive best practices for CAPS data labeling, which is crucial for supervised learning methods that are considered helpful. Privacy compliant behavior is analyzed in this thesis. Vehicle HMI are therefore analyzed with regard to privacy concerns and regulations. The user's data protection perspective is also captured in a survey. This is necessary to find indications that CAPS is not only an alternative from a technical point of view. It is also an alternative that could be in the user's favor.
ThesisNote
Darmstadt, TU, Diss., 2021
Beteiligt
Verlagsort
Darmstadt