Aiming at the problem that the pilot aided channel estimation of OFDM system in fast fading environment is greatly affected by channel state fluctuation, a channel identification and compensation method for OFDM system based on artificial neural network is proposed. Firstly, the traditional decision feedback channel estimation method is analyzed, and its problems in fast fading environment are explained. Secondly, the decision feedback channel estimation is used to obtain part of the channel state information at constant intervals, so that only a small amount of estimated channel state information is used to train the artificial neural network. Then, Levenberg-Marquardt algorithm is used for neural network training. Finally, after the training of the artificial neural network, all data symbol indexes are serially input to the artificial neural network, so as to interpolate the whole transition of the channel state information, thus effectively compensating the channel variation. The test results of OFDM communication system in fast fading environment show that, compared with traditional estimation methods, this method has better BER performance in high mobility environment with Doppler frequency of 700 Hz, and can eliminate error flat layer, and the BER is below 10-4.