关键词: 深度学习, OFDM, 信道估计, SRWGAN

Abstract: In the channel estimation process of the orthogonal frequency division multiplexing (OFDM) system, the traditional channel interpolation algorithm is based on the assumption that the estimated values near the pilot are correlated. When the channel characteristics are discontinuous due to the time-varying and frequency-varying characteristics of the wireless channel, the estimation results will be unsatisfactory. In response to this problem, this paper introduces an improved model SRWGAN of super-resolution reconstruction model SRGAN to replace the interpolation processing in channel estimation. In the model SRWGAN, the least squares (LS) estimation value at the pilot is analogous to the pixels in the low-resolution image. The channel features are first extracted through the convolutional network, and then the mapping relationship is learned through multiple residual networks. Then it is amplified by the up-sampling layer, and finally the discriminant network WGAN is used to continuously discriminate and improve the estimation effect. The experimental results show that the channel estimation effect based on SRWGAN is better than the traditional channel estimation algorithm, and compared with the same type of SRCNN model, under the same conditions, when the bit error rate is the same, the signal-to-noise ratio (SNR) is improved by about 3 dB, and when the MSE value is the same, the SNR is increased by about 5 dB.

Key words: deep learning, OFDM, channel estimation, SRWGAN