A spatial domain multiplexing technology for fiber specklegram sensor.

• A method to solve the lack of multiplexed ability for fiber sensor based on specklegrams. • A specklegram can be decomposed spatially according to different incident angle. • The annular specklegram can be formed by the incident of side-coupling light. • Transfer learning technology is employed to improve the accuracy of multiplexed sensor based on specklegram recognition. With the advancement of image processing and artificial intelligence technology, Fiber Specklegram Sensors (FSS) based on specklegram recognition have emerged as a novel sensing technology. The output speckles represent the interference between different transmission modes in the fiber, which vary with changes in fiber fluctuations such as the pressure on the fiber or ambient temperature. This sensor boasts high precision and low cost. However, when the optical fiber is required to perform multi-parameter or multi-position sensing simultaneously, the variety of specklegrams counterpart to external stimuli has become very large, significantly limiting the application scenarios of FSS. In this paper, a new structure based on side-coupling theory is proposed, where the output specklegram comprises a circular specklegram and an annular specklegram formed by end-coupling light and side-coupling light respectively. In this manner, the circular specklegram and annular specklegram can represent different sensing parameters, and the circular specklegram is completely unaffected by the annular specklegram, enabling multiplexed sensing. Multiplexed experiments for pressure and temperature sensing are conducted utilizing this structure, and deep learning technology is employed to process the different sets of specklegrams for regression prediction. The accuracy of pressure and temperature sensing tests reached 98.62% and 95.31% respectively, verifying the feasibility of this method for multiplexed sensing. [ABSTRACT FROM AUTHOR]