It is known that most subspace identification algorithms give biased estimates for closed-loop data due to a projection performed in the algorithms. In this work, consistency analysis of SIMPCA is given and the exact input requirement is formulated. The effect of column weighting in subspace identification algorithms is discussed and the column weighting for SIMPCA is designed which gives consi...

This paper advocates a new subspace system identification algorithm for the errorsin-variables (EIV) state space model via the EM algorithm. To initialize the EM algorithm an initial estimate is obtained by the errors-in-variables subspace system identification method: EIV-MOESP (Chou et al. [1997]) and EIV-N4SID (Gustafsson [2001]). The EM algorithm is an algorithm to compute the maximum value...

In this paper, the blind channel identification problem is formulated in a stochastic state space framework. Starting from a state space model we present a preprocessing step based on two orthogonal subspace projections. Using these orthogonal projections, we derive an algorithm for blind channel estimation which is insensitive to the spatial color of the noise. The performance of this new algo...

We illustrate procedures to identify a state-space representation of a passive or bounded-real system from noisefree measurements. The basic idea underlying our algorithms is to obtain a state sequence from a rank-revealing factorization of a Gramian-like matrix constructed from the data. The computation of state-space equations is then performed solving a system of linear equations, similarly ...

A sequence ${T_n}_{n=1}^{infty}$ of bounded linear  operators on a separable infinite dimensional Hilbert space $mathcal{H}$ is called subspace-diskcyclic with respect to the closed subspace $Msubseteq mathcal{H},$ if there exists a vector $xin mathcal{H}$ such that the disk-scaled orbit ${alpha T_n x: nin mathbb{N}, alpha inmathbb{C}, | alpha | leq 1}cap M$ is dense in $M$. The goal of t...

ABSTRACT This paper presents two subspace-based methods, from the MOESP (MIMO output-error state space) family, for state-space identification of continuous-time fractional commensurate models from sampled input-output data. The methodology used in this paper involves a continuous-time fractional operator allowing to reformulate the problem so that the state-space matrices can be estimated with...

Tony Robinson Cambridge University Engineering Department Cambridge CB2 IPZ England [email protected] The speech waveform can be modelled as a piecewise-stationary linear stochastic state space system, and its parameters can be estimated using an expectation-maximisation (EM) algorithm. One problem is the initialisation of the EM algorithm. Standard initialisation schemes can lead to poor forma...