Aiming at the shortage of tradition prediction method, established a new prediction model of gyro drift based on wavelet neural network. This paper mainly studies recognition and abstraction method of gyro output signal’s drift trend using wavelet analysis theory, set up the non-linear prediction model of drift signal combining the wavelet neural network, and predicts the drift trend taking certain gyro measured signal as example. Result shows that the model can reduce human factor influence, advance drift prediction effect and navigation system precision.