Signal Classification Based on Analog Computing Using MEMS Network

The rising complexity of machine learning algorithms and Artificial Intelligence in many applications, such as smart building, has prompted the development of alternate computing options. Because of their compact size, low power consumption, and diverse functionality, microelectromechanical systems (MEMS) have emerged as a possible candidate. This thesis focuses on using MEMS networks as computing units to classify a simple signal classification task using neural network methodology. The study intends to show the potential of using MEMS as an analog computing unit by discussing the advantage of the bi-stability pull-in behavior and hysteresis to create an accurate classifier of these waveforms. Modeling and simulation are being conducted to assess the MEMS-based computer units performance. The results reveal that the proposed methodology performs the required classification without requiring a digital computer. Furthermore, This study adds to the field of analog computing with MEMS by providing insights into the feasibility and potential of using MEMS networks for more complex classification tasks such as those related to smart building applications. Adviser: Fadi Alsalee