Identification of stationary linear dynamic systems multivariate on input

In article the problem of an estimation of parameters linear difference equations with a multivariate input is considered at presence of handicapes of supervision in input and output signals. This task differs from a standard problem regression estimation, the new criterion of estimation is offered on the basis of the relation of two square-law forms, generalizing standard method of the least squares and allowing to receive wellgrounded estimations of parameters. The numerical method of definition of estimations of parameters linear difference the equations, reduced to the repeated decision linear difference the equations is offered also.