Porting an AI Powered Wearable Health Monitor to Zephyr on Open Hardware - Szymon & Jakub

Porting an AI Powered Wearable Health Monitor to Zephyr on Open Hardware - Szymon Duchniewicz, Avanade & Jakub Duchniewicz, Tietoevry

To RTOS or not to RTOS? Szymon and Jakub will introduce obstacles they faced and decisions behind moving a closed-source single-threaded wearable health monitor to an RTOS, open-hardware based system with an AI model deployed on a Field-Programmable Gate Array (FPGA). They will share tips on how to get started with Zephyr development, when and why to build a system using an RTOS. They will also share best practices and experiences on deploying a Machine Learning model to an embedded FPGA and interacting with it from Zephyr OS. The project dissected in this talk uses QuickLogic's and Antmicro's QuickFeather board, powered by Open Hardware EOS S3 System on Chip. The Machine Learning Model is deployed using TensorFlow Lite and then integrated using an open source FPGA toolchain. Device collects data from an SPO2 sensor that is parsed by Zephyr at runtime and finally passed to the model deployed on the FPGA for inference. The results are then returned to the Zephyr RTOS and displayed on a small OLED screen. The model is trained using open source data pertaining to blood pressure estimation based on SPO2 levels and is tailored specifically for deployment in embedded scenarios.