A Vital-Signs Monitoring Wristband With Real-Time In-Sensor Data Analysis Using Very Low-Hardware Resources

We present a cost-effective multisensor wristband that monitors vital signs, including heart rate (HR), respiratory rate, and oxygen saturation, with very low-hardware resources, while effectively mitigating motion artifacts. Our system, which operates in real time, optimizes resource utilization and minimizes memory footprint by intelligently managing motion artifacts during vital signs extraction. For this purpose, we introduce a new intelligent domain fusion (IDF) algorithm, which allows to reduce resource allocation during moderate motion artifacts by up to 30% in memory usage, while achieving an average latency of as short as 18 ms in worst case. To further enhance efficiency, our frequency-domain analysis (FDA) is optimized through the utilization of a Z-chirp transform, minimizing memory utilization during Fourier transforms. Our prototype is built around a real-time operating system (freeRTOS) running on an ARM-M4 architecture, enabling Bluetooth lowenergy (BLE) 5.0 control, data processing, and communication. It incorporates an onboard 1-Gbit Nand flash memory for continuous data storage, when no wireless connection is available, offering up to 7 h of uninterrupted sensor data recording. The sensor data collected by our prototype encompass a comprehensive range of raw and processed information critical for monitoring various physiological and environmental parameters in a concise message format comprising a data ID, a data payload, and a timestamp. With a small 250 mAh battery, our prototype provides an autonomy of 18 h of continuous data collection, averaging a mere 10.41 mAh of power consumption.