In order to solve the problem that noise and signal interference have an effect on the channel estimation of receiver, the channel estimation method based on integrated neural network has been proposed in this thesis. Firstly, improved neural network will be used to extract nonlinear relationship between pilot signal with noise and original pilot signal. Then the diversity factor will be calculated based on its output errors, and the integrated neural network model will be obtained according to diversity factor. Finally, channel estimation results can be obtained by model. Compared with the LS estimation, not only can this method improve the reliability of the communication, but also reduce the cost of pilot to improve the effectiveness of communication. Under different modulation methods, when the bit error rate is the same, the signal-to-noise ratio (SNR) of LS algorithm is lower; based on orthogonal frequency division multiple (OFDM), the signal-to-noise ratio (SNR) of LS algorithm is lower. The simulation results of OFDM communication system show that the method has good performance.