Deep learning based force recognition using the specklegrams from multimode fiber

Abstract

The force induced variations of interferences in multimode fiber (MMF) are recognized by the output specklegrams. In this work, the classification of specklegrams is reported to identify the magnitude and position of the force applied on the MMF. The specklegrams from the MMF are recorded by a CCD camera at different force conditions. Because of the large number of transverse modes in the fiber, the specklegrams contains abundant information about the force applied on fiber states. By employing a convolutional neural network (CNN), the classification accuracies of the force position and magnitude on the fiber were 95.91% and 96.67% for test dataset. This reported scheme has the advantages of low cost and simple structure and is suitable to identify specific types of force in distributed sensing applications.

Keywords:

• Convolutional neural networks (CNN)
• deep learning
• fiber specklegrams
• force recognition
• multimode fiber

Disclosure statement

No potential conflicts of interest are reported by the authors.

Additional information

Funding

Our work is supported by the National Natural Science Foundation of China projects (Nos. 62075132 and by Natural Science Foundation of Shanghai (22ZR1443100).