In orthogonal frequency division multiplexing systems,the pilot used for channel estimation occupies valuable transmission resources and consumes user equipment energy. To tackle this issue,a channel estimation method combining differential detection and deep neural network is proposed. At the transmitter,the transmitted data are differen-tially encoded. At the receiver,according to the idea of decision-directed channel estimation,the recovered data with dif-ferential decoding are regarded as the transmitted pilot to capture the initial features of the channel estimation. With the help of the captured initial features,an enhanced channel estimation network (En-CENet) is built to improve the channel estimation accuracy by integrating the differential features and channel features captured by the neural network. The simu-lation results show that,compared with the pilot-based channel estimation method and machine learning superposition channel estimation method,the proposed method improves the channel estimation accuracy while improving the spectral efficiency,saving the energy consumption of user equipment and reducing the computational complexity and running time of receiver.
维普
万方数据