A Fiber-Optic Sensor-Embedded and Machine Learning Assisted Smart Helmet for Multi-Variable Blunt Force Impact Sensing in Real Time

Abstract

Early on-site diagnosis of mild traumatic brain injury (mTBI) will provide the best guidance for clinical practice. However, existing methods and sensors cannot provide sufficiently detailed physical information related to the blunt force impact. In the present work, a smart helmet with a single embedded fiber Bragg grating (FBG) sensor is developed, which can monitor complex blunt force impact events in real time under both wired and wireless modes. The transient oscillatory signal “fingerprint” can specifically reflect the impact-caused physical deformation of the local helmet structure. By combination with machine learning algorithms, the unknown transient impact can be recognized quickly and accurately in terms of impact magnitude, direction, and latitude. Optimization of the training dataset was also validated, and the boosted ML models, such as the S-SVM+ and S-IBK+, are able to predict accurately with complex databases. Thus, the ML-FBG smart helmet system developed by this work may become a crucial intervention alternative during a traumatic brain injury event.