Coursera Spring Sale
40% Off Coursera Plus Annual!
Grab it
Explore the development of a revolutionary textile-integrated breath sensor that transforms ordinary clothing into real-time health monitoring devices in this 14-minute conference presentation. Discover how Georgios Kokkinis and his team at Silicon Austria Labs engineered a working prototype that combines embroidered interconnects, 3D-printed dielectric islands, and carbonized-silicon yarn strain gauges to create truly wearable healthcare technology. Learn about the critical challenges in building flexible, durable interconnects for wearables and understand why traditional peak detection methods fail under real-world conditions involving drift, burn-in, and noise. Examine the team's approach to creating custom datasets with clean ground truth data and compare CNN versus DNN architectures using Edge Impulse's EON Tuner for optimal performance. Gain insights into the deployment process on STM32L4 MCU systems utilizing DMA, timers, and CMSIS-DSP preprocessing techniques. Analyze the trade-offs between model accuracy, speed, and power consumption, with CNN models proving more robust despite slower performance compared to faster but less resilient DNN alternatives. Understand the importance of sensor-specific datasets and fine-tuning for smart textile applications, and explore future developments including dataset expansion, cross-garment generalization improvements, and potential hydration sensing capabilities for seamless, private, all-day remote patient monitoring.
