Edge Computing Device with Embedded Intelligence for Health Status Classification

Advances in artificial intelligence and embedded systems processing capabilities have contributed to monitoring people's health status, enabling the development of wearable devices and edge computing solutions. However, there are problems related to accuracy, energy efficiency, and signal transmission in embedded devices, which limit their application in health monitoring systems. This paper describes the development of an ESP32-based hardware device with health classification algorithms using embedded integration techniques and artificial intelligence (TinyML). The hardware device is developed by integrating temperature, heart rate, SpO2, and respiratory rate sensors. During the evaluations, it was found that the decision tree and random forest models offered accuracy greater than 85%, while the Support Vector Machine model presented greater memory consumption and sensitivity to variations in SPO2. Moreover, the random forest model had the highest accuracy (0.975) in both the software and the edge device, highlighting the importance of balancing efficiency and inference time.