基于递进感知Transformer的通感一体化链路损伤补偿方法OA

To achieve low-impairment reception and demodulation of communication signals in an integrated sensing and com-munication fiber link,this paper proposes a link impairment compensation method based on a progressive perception Trans-former.By limiting the receptive field of the attention mechanism,the proposed progressive perception Transformer model fo-cuses on the neighborhood information of each symbol.The receptive field is increased layer by layer to emulate the formation and accumulation process of physical impairments in the actual link,thereby compensating for link signal impairments more ef-fectively.Experiments are conducted on a 25 km fiber integrated sensing and communication link transmitting a 4-level pulse amplitude modulation(PAM-4)signal in an access network scenario.The experimental results show that,compared with the decision-directed least mean square method,the progressive perception Transformer model achieves a received optical power penalty reduction of more than 2 dB at a bit error rate of 10⁻².