A parameter identification method is presented that identifies all the parameters of an induction motor simultaneously. The method assumes that the motor can be described by a time-varying linear model. The idea is to filter the current, voltage and speed signals of the motor such that a set of signals results that are related by simple linear equations. The coefficients of these linear equations are the same as the coefficients of the differential equations of the machine. Two estimation methods are investigated: the first tries to identify the parameters with a normal general total least squares method, while the second uses constrained general total least squares. The second method is shown to be more robust than the first with respect to noise sensitivity, which is important since the system must function in an industrial environment. The problem of identifying induction motors is a very difficult one, because interesting phenomena like pole-zero cancellations and non-persistency of excitation may occur.