Numerical Analysis of the Estimate of Noise Covariance by using Best Linear Unbiased Estimate.

This paper derives a BLUE (best linear unbiased estimate) for the estimation of noise covariance matrix in the framework of subspace identification. An estimation method of the noise covariance with a CCA (canonical correlation analysis) weighting matrix was previously proposed by the authors. The newly derived BLUE is compared with the previously proposed method numerically and it is shown that the previously proposed method gives a good performance.