A Novel Mems Reservoir Computing Approach for Classifying Human Acceleration Activity Signal
Mohammad Okour, Mohammad Megdadi, Mutaz Al Fayad, Abdallah Al Zubi, Sulaiman Mohaidat, Fadi Alsaleem · 2025
Abstract Neuromorphic computing, drawing inspiration from the human brain, harnesses specialized hardware and software to mimic intricate information processing. A pivotal component within this domain is the Micro-Electro-Mechanical Systems (MEMS). This paper marks an effort by introducing a MEMS reservoir computing model that departs from conventional virtual node concepts. This novel approach couples multiple MEMS systems to create dynamic and high-dimensional responses. The primary objective of our study is to distinguish between walking and running signals based on acceleration measurements. Our research advances the boundaries of reservoir computing and MEMS applications and marks an important milestone in signal processing analysis and classification. In this paper, we achieved a classification accuracy of 77%, demonstrating the practical potential of this technology across various applications in wearable technology and beyond.