The locally stationary parametric model was established to analyze the nonstationary random vibration environment of aircraft during lift-off. The computation of a time-varying spectrum was decomposed into two separate operations, one to define the time-varying mean square value of the signal, and the other to define the spectral properties of the signal. The wavelet transforms were used to separate the nonstationary vibration signal into the deterministic signal and random signal. Specifically, the time-varying mean square value of the signal was computed using the entire signal bandwidth,and the averaging time could be made short without causing a large bias error. Conversely, the basic spectra character of the signal was computed using the entire signal duration, and the frequency resolution bandwidth could be made small without causing a large random error. The vibration environment of aircraft during lift-off was obtained by concluding the auto-spectra of stationary random vibration.