15 de noviembre de 2024
Resumen:
Over the past decade, the irruption of the Internet of Things and Point-of-Care diagnostics has increased the demand for low-cost (bio)sensors. At the same time, this trend has evolved in parallel to the growing field of Machine Learning (ML), and therefore, this PhD research addresses the inevitable convergence of both technologies, focusing on microwave resonant sensors as the key technology for developing low-cost sensors. Although this sensor type has the potential to be both cost-effective and highly sensitive, due to the lack of specificity, its performance can be hindered by environmental noise or complex applications with many analytes involved. Therefore, this PhD research aims to leverage ML techniques to enhance the capacitive sensing capabilities of MW resonant structures, thereby developing low-cost, high-performance sensors. In order to fulfill this objective, the corpus of this PhD thesis consists of three published papers and the preliminary results of a fourth work. Together, this work proves that implementing ML models enables the substitution of complex instrumentation for low-cost electronics and open hardware. In addition, they demonstrate how the trained models with sensor signals can help investigate the sensing principle's nature and improve the interpretation of the results. Finally, this research investigates how Deep Learning models can generate realistic synthetic data that could increase the size and quality of the acquisition dataset, as well as allow the interpolation of undetected signals.
Descriptores: Dispositivos de Microondas, Magnitudes Eléctricas y Su Medida, Equipo de laboratorio, Aparatos Científicos, Inteligencia Artificial, Diseño de Sistemas Sensores
Cita:
M. Monteagudo Honrubia (2024), Machine Learning techniques to enhance the capacitive sensing of microwave resonant structures. Madrid (España).
