Wireless Tag Sensor Network for Apnea Detection and Posture Recognition Using LSTM

Abstract

We have developed a low-cost, high-accuracy, and energy-efficient wearable tag sensor for apnea detection. The sensor can detect different types of breathing problems by monitoring the small movements of the chest wall compartments during each respiration cycle. This tag sensor sends also apnea events, digital respiration rate, and patient posture data using an ultra high radio frequency identification (UHF RFID) reader. The reader is based on the recent AS3993 chip connected to a Raspberry Pi 4 controller, which acts as a local server and is connected to the cloud to share acquired data with the treating doctor. A sleep disorder detection and classification with several positions using a long short-term memory (LSTM) network algorithm is implemented in real-time on the embedded arm microcontroller STM32F407. The proposed apnea detection method exhibits low error, enabling it to meet clinical requirements. The accuracy of apnea events and position detection were triggered in over 93% of cases. We have also evaluated six different classification techniques optimized by considering the proposed feature extraction and regularization of classifier parameters.