Cover story: Zheng C H, Wang T S, Liu Z Q, et al. Deep transfer learning method to identify orbital angular momentum beams[J]. Opto-Electron Eng, 2022, 49(6): 210409

Orbital angular momentum beams was proposed by Allen in 1992, which provides a new multiplexing method in addition to wavelength/frequency, time, space and polarization state for the field of optical communication. In addition to using the orbital angular momentum beam as the carrier, it can also be encoded according to the unique light intensity distribution, enabling information transmission. 

The research team of Professor Wang Tianshu from the National and Local Joint Engineering Research Center of Space Optoelectronic Technology of Changchun University of Science and Technology proposed to use the transfer learning method of deep learning to identify orbital angular momentum beams. Transfer learning refers to knowledge transfer in similar but different domains or tasks. In order to simulate the atmospheric turbulence, we generate the atmospheric turbulence phase screen by the subharmonic method and build the simulated turbulence environment by loading the phase screen on the spatial light modulator.The training speed of the transfer learning model is 2.3 times faster than the traditional recognition model. It is proved that the transfer learning system can significantly reduce the time required for model training while maintaining a high recognition rate. It provides a new solution for the future's rapid construction of orbital angular momentum identification system.