Simulation results show that non-stationary noise can be identified in time domain and resolution improvement is obvious compared with that of conventional noise detection in frequency domain.

By exploiting wavelet packet transform to analyze signals both in time and frequency space, this paper researches the statistic property of non-Gaussian stationary noise and its signal structure.

The shock wave and strong pulse noise, which produced by light rocket launcher, are belong to a non stationary signal process with a wide frequencies.