Thin fabric pressure sensors and TinyML smart gloves for edge IoT
TL;DR
This study presents a smart glove integrating thin carbon nanotube-fabric pressure sensors with embedded tiny machine learning (TinyML) models on an ESP32S3 platform, demonstrating a compact and privacy-aware wearable platform for real-time human-machine interaction and embedded healthcare systems.
Abstract
Summary Wearable human-machine interfaces require flexible sensing, low-latency processing, and secure data handling for practical edge IoT applications. This study presents a smart glove integrating thin carbon nanotube (CNT)-fabric pressure sensors with embedded tiny machine learning (TinyML) models on an ESP32S3 platform. The sensor achieves a sensitivity of 0.46 kPa−1, a response/recovery time of 60/45 ms, and stable operation over 3,500 cycles. Pressure signals from three fingertip channels are processed locally using lightweight random forest and k-nearest neighbors models, achieving classification accuracies of 92.1% and 96.55% with inference times of 0.185 and 4 ms, respectively. All raw sensor data remain on the embedded device, while only prediction labels are transmitted through Wi-Fi for remote monitoring. The proposed framework demonstrates a compact and privacy-aware wearable platform for real-time human-machine interaction and embedded healthcare systems.