Event-based Speckle Interrogation for High-Bandwidth Multi-point Optical Fiber Sensing

Abstract

Speckle-based fiber optic sensors are well-known to offer high sensitivity but are strongly limited on the interrogation side by low camera frame rates and dynamic range. To address this limitation, we present a novel interrogation framework that explores event-based vision to achieve high throughput, high bandwidth, and low-latency speckle analysis of a multimode optical fiber sensor. In addition, leveraging a tensor-based decomposition of the raw event streams through multi-point calibration and machine-learning optimization, our approach also proves capable of isolating simultaneous deformations applied at distinct points. The experimental results validate the methodology by separating the signals of four piezoelectric actuators over a 400Hz-20kHz range with minimal crosstalk applied over varying distances from 3cm to 75cm. Finally, extending the impact of the work with an acoustic sensing proof-of-concept, we have coupled the fiber to two plastic enclosures and recovered separable audio signals between 400 and 1.8 kHz with minimal waveform distortion. Overall, these results establish event-driven speckle interrogation as a new versatile platform for real-time, multi-point acoustic sensing and pave for its application in complex and unstructured environments in future works.