Novel algorithms based on conjunction of the Frisch scheme and extended compensated least squares

The paper presents a general framework for the Frisch scheme and the extended compensated least squares technique within which two new algorithms for the identification of single-input single-output linear time-invariant errors-in-variables models are proposed. The first algorithm is essentially the Frisch scheme using a novel model selection criterion. The second method is a modification of the extended compensated least squares technique, which utilizes not only the set of overdetermined normal equations, but also the Frisch equation to solve the parameter estimation problem. An extensive Monte-Carlo simulation compares the novel algorithms with existing errors-in-variables identification approaches.