A residual self-tuning RELS algorithm is proposed to estimate ARMA parameters. Pseudo white noise was introduced to compensate for the error contributed by the unknown noise. DDS and AIC criterion are used for order selection.

• 提出了残差自校正的RELS算法,在ARMA模型参数估计的过程中引入虚拟白噪声,来补偿参数估计过程中的由于噪声未知而引入的误差,并将DDS法与AIC准则相结合确定模型阶数。