Frequency Domain Identiication, Subspace Methods and Periodic Excitation

Recent frequency domain identiication algorithms based on subspace based techniques are discussed. The algorithms construct a state-space model by means of extraction of the signal subspace from a matrix constructed from frequency data. A singular value decomposition plays a key part in the subspace extraction. The subspace methods are non-iterative methods in contrast to classical iterative parametric optimization techniques. The use of periodic excitation leads to a leakage free discrete Fourier transform of the measured data as well as simple noise reduction possibilities by averaging.