Wearable AI

The design and realization of wearable devices differs significantly from standard mobile devices. Touch interaction is often impossible, sensors and other components might be distributed and the software design is shared over several nodes. Special attention here has to be paid to the realization of hardware and software components. Especially aspects of power management, the design of software architectures, the implementation of actuators and the integration into textiles is of major importance.

Recent development in this area has shown that mobile applications from smartphones cannot directly be transferred to embedded textile systems. A specialized design concerning seamless integration and novel rugged, non-obtrusive components, even for underwater usage, have to be developed considering aspects of user acceptance.

The ability to place devices in close contact with different body parts opens up new sensing concepts: from new ways of measuring vital signs through the detection of emotions to various internal body function. Such sensing concepts must combine the ability to reliably extract useful information with

a variety of wearability constraints such as flexibility, weight, power consumption,tolerance of motion artifacts and price. The design, implementation and testing of novel wearable sensors tailored to the above requirements is a key competence at DFKI. Previous work includes new textile capacitive sensors for nutrition monitoring, novel magnetic motion tracking sensors and large scale, textile restive pressure sensor matrices.

The realization of embedded distributed sensor and actuator frameworks naturally generates a huge amount of data. This data has to be analyzed and processed in order to provide interactive feedback. In comparison to standard desktop applications special attention has to be paid on the mobile nature of wearable technologies. Power availability, processing capabilities and storage are limited resources, especially if long term operation is considered. Topics of interest are robust activity recognition, personalized data processing, power-aware sensor data fusion and user modeling. Physical and mental health monitoring and different types of personal assistance systems are important applications.

Data analysis tools for wearable devices additionally have to deal with temporal sensor loss, human acceptance and privacy issues.

Wearable devices by nature are collecting information about the user him- or herself. Motion data, localization information or vital signs are of importance, just to mention a few. If online feedback is required a novel interaction channel for embedded devices has to be developed. Although several techniques like head-up displays are available the acceptance of those components is often very low. Research topics include therefore non-intrusive communication channels for immediate real-time feedback, such as haptic feedback via tactile actuators distributed over the body or audio feedback.

Due to the large amount of data and the resulting information from the data analysis steps special focus has to be laid on usability, interaction design and context-aware interfaces. In contrast to desktop applications more focus on multimodal interfaces is required.