According to the practical problem of GPS/SINS integrated navigation system encountered, combine the nonlinear prediction algorithm of artificial neural network (ANN) with genetic algorithm to establish the SINS error feedback correction algorithm based on genetic wavelet ANN. Research on the wavelet ANN learning algorithm based on genetic algorithm, and confirm its structure model. When GPS signal can be used, acquire ANN training sample according to GPS/SINS integrated navigation input and output signal, carry out online ANN training, and acquire optimal ANN model parameter. When GPS signal is broken up, according to trained ANN module, predict SINS position error, speed error and gesture error of GPS signal unlocking. Then correct SINS error to acquire correct navigation parameter. The simulation test results show that the algorithm can effectively improve the precision of integrated navigation system when GPS observation data is unreliable.