Tapered Optical Fiber Enabled Distributed Sensors with High Spatial Resolution by Deep Learning

Distributed optical fiber sensing (DOFS) systems are valuable tools for monitoring various physical parameters (e.g., temperature, pressure, strain). DOFS systems, however, remain a challenge for achieving high spatial resolution in narrow spaces due to the bulky external demodulation techniques. Herein, by taking advantage of spatial inhomogeneity-induced higher-order modes, a tapered optical fiber-based distributed sensor is developed to monitor the contact position and force by deep learning simultaneously. The sensor achieves the position and force prediction resolutions of 7.6 mu m and 0.02 N, respectively, over the 5.6 mm length by developing a multi-scale convolutional neural network with a long short-term memory model to decode the output signals. Furthermore, the sensor can identify position and force in a 2D plane, exhibiting excellent distributed sensing capabilities. The results may pave the way toward high-performance distributed sensors for applications from healthcare, robotics to human-machine interfaces. Tapered optical fiber-based distributed sensor, which can simultaneously monitor the contact position and force, is developed by taking advantage of spatial inhomogeneity-induced higher-order modes. The sensor achieves position and force prediction resolutions of 7.6 mu m and 0.02 N, respectively, by deep learning algorithm. Furthermore, the sensor can identify position and force in a 2D plane, exhibiting excellent distributed sensing capabilities. image