Adaptive identification of a modified ARMA lattice model

An adaptive method which identifies a modified ARMA (autoregressive moving average) lattice model is proposed. The modified model is designed as a compact structure with a limited number of parameters. The method consists of time and order-update recursions for model identification, an estimate of a self-tuning input, and a stability check on an estimation model. In addition, the method employs a weighting factor to adjust a window length to observed data. Compared with other ARMA lattice models and their identification methods, the proposed method has several interesting features. In particular, it is possible to use arbitrary AR and MA orders for the estimation model. The proper orders are quite suitable for identifying a complicated structure. The estimation model can be realized using fewer parameters than the conventional ARMA lattice models. Since a limited number of parameters are estimated, low calculation cost is expected. The proposed method can be applied to the ARMA models with unknown inputs