Optimizing NFC-Based Wearable Sensors for Arterial Pulse Monitoring: A Comparative Study of Sampling Rates and Machine Learning Models

This paper details the development and optimization of a novel NFC-based wearable pressure sensor designed for arterial pulse monitoring with a focus on energy efficiency and data transmission management. By integrating this sensor with various machine learning models for human activity recognition using arterial pulse signals, we evaluate their impact on optimizing the sampling rate. This optimization directly influences power consumption and data transmission rates. A comparative analysis reveals that the 1-dimensional Convolutional Neural Network (1D-CNN) maintains an accuracy of over 92% while allowing the sampling rate to be lowered from 100 Hz to 40 Hz. This reduction leads to a 35% decrease in power consumption and a 50% decrease in data transmission needs. In contrast, traditional models such as Random Forest (RF) and Support Vector Machine (SVM) require a minimum of 70 Hz to maintain comparable accuracy, achieving only a 19% reduction in power and a 30% reduction in data transmission rates. Our findings emphasize the critical role of model selection in optimizing the arterial pulse sampling rate, enhancing the energy efficiency and operational effectiveness of our wearable health monitoring device.