We would like to congratulate our colleague, Severin Bernhart, on successfully completing his doctorate! The data engineer from the Human Motion Analytics research group was awarded the Doctor of Technical Sciences degree on 4 December 2025. His thesis defense took place on 27 November 2025 at the Faculty of Digital and Analytical Sciences at Paris Lodron University of Salzburg.

Doctoral thesis on edge AI in smart textiles for health-promoting running

In his dissertation at the Department of Artificial Intelligence and Human Interfaces’ Computer Science department, Severin Bernhart focused on the potential of smart textiles to support healthy breathing patterns while running. Under the title “Bra4Vit: Breathing Guidance Apparel in Running via Edge-AI in Smart Textiles,” Bernhart developed an innovative system that uses textile-integrated sensors and edge AI to monitor and support runners’ breathing in real time.

Abstract of the dissertation

Running is a widely practiced form of physical activity due to its low barriers to entry. However, breathing difficulties can limit participation and reduce enjoyment of exercise. Continuous recording of breathing patterns and the provision of real-time feedback enable the detection of irregularities and the targeted influencing of breathing behavior. Rhythmic breathing with a prolonged exhalation activates the parasympathetic nervous system, promoting relaxation and increasing motivation. Vitality, as an essential component of well-being, can thus be positively influenced by physiological regulation and regular exercise. Smart textiles offer great potential for providing respiratory monitoring and guidance in the field. However, challenges remain, particularly in developing robust sensor technologies, integrating conductive yarns, and standardizing electronic interfaces. Conventional AI-supported methods of respiratory analysis often rely on centralized high-performance computing capacity, which hinders the provision of time-critical feedback. Edge AI enables local processing on the device, ensuring latency-free feedback. This work aims to develop a personalized breathing guidance system for female runners, supporting rhythmic, controlled breathing and prolonged exhalation. A comprehensive literature review of body networks, respiratory sensors, feedback strategies, and low-latency signal processing informed the development of Bra4Vit: a smart bra for vitality with a textile-based respiratory sensor and a miniaturized on-body unit incorporating an inertial sensor for step detection. The developed Edge AI model recognizes inhalations and exhalations, breathing phases, and continuous breathing segments with accuracies of 93.1%, 85.6%, and 81.5%, respectively, without any time delay. Chain knitting technology and an individual chip design optimize wearing comfort, handling, and production costs. However, Bra4Vit currently lacks sweat resistance, a feature that will be implemented in future work. Planned field evaluations will assess the accuracy, robustness, comfort, and intuitive usability of the breathing guide.

Severin Bernhart’s research is a valuable contribution to the development of wearable, AI-based systems designed to support athletes with their breathing techniques, with the ultimate goal of improving motivation, vitality, and long-term health.

Salzburg Research would like to extend its warmest congratulations on this outstanding scientific achievement and looks forward to continuing its collaborative research and development projects.