Abstract： Under the hypothetical condition that the stochastic process is non-Gauss, stationary and statistic independence and seismic wavelet is non-causal and mixed phase, the paper carried out the study of wavelet pickup and model adaptability by appli2 cation of moving average (MA) and autoregressive moving average ( ARMA ) models to create the models of seismic records respectively and using high-order-cumulants-based linear solution approach —cumulants matrix equation approach that is small operational efforts. The results of numeric simulation and real seismic data processing showed that the ARMA model was characterized by parametereconomic and high efficiency in comparison with MA model ; the cumulants matrix equation approach can more effectively suppress the addictive Gauss colored noise but requires the estimation of cumulants samples to have high accuracy ; if the errors of estimation of cumulants samples and variance are appropriate, the described cumulants matrix equation approach combing with ARMA model can highefficiently and accurately estimate the seismic wavelet.
