Identification of Errors-in-Variables Systems via Extended Compensated Least Squares for the Case of Coloured Output Noise

The paper deals with the identification of dynamic discrete-time linear time-invariant errors-in-variables systems for the case of coloured output noise. The proposed algorithm is constructed within an extended bias compensated least squares framework, where the principle of the bilinear parametrisation is exploited. Two standard least squares methods are utilised to separately determine, firstly, the system parameter vector and, secondly, the input measurement noise variance together with auto-correlation elements of the output noise. Since it is based on the least squares technique, the computational complexity of the algorithm is relatively low. A Monte-Carlo simulation compares the developed technique with other errors-in-variables identification methods.