Intelligent automatic modulation recognition for OFDM with index modulation in underwater OWC systems

Orthogonal frequency division multiplexing with index modulation (OFDM-IM) is shown to be an efficient modulation scheme for underwater optical wireless communication (UOWC) systems due to its high power efficiency and excellent flexibility. Differing from conventional OFDM, the modulation parameters of OFDM-IM consist of not only M -ary constellations but also subcarrier indices. In this Letter, we propose and demonstrate an intelligent automatic modulation recognition (AMR) scheme for OFDM-IM in UOWC systems by using Swin Transformer (Swin-T). Moreover, transfer learning with a hybrid training approach is further proposed to substantially enhance the AMR performance. Experimental results successfully demonstrate the feasibility and superiority of the proposed intelligent AMR scheme based on Swin-T with transfer learning and hybrid training in UOWC systems. More specifically, a recognition accuracy of 96% can be obtained by using Swin-T with transfer learning and hybrid training for a relatively small received signal-to-noise ratio (SNR) of 3.6 dB.