| 摘要 |
The prior art knows two different approaches to calibration of multichannel instruments, viz., the so-called physical and statistical calibration methods. The new methods translate the difficult inverse-problem posed by the statistical method into simpler, forward-problem, "physical" measurements of the signal and the noise. The new methods combine the quality of the statistical method with the low cost and interpretability of the physical method. The new methods disclose how to compute the optimal regression vector; how to update the optimal regression vector to account for small changes in the noise; how to choose a "good" subset of channels for measurement; and how to quantify the noise contributions from the multichannel measurement and from the reference measurement individually. The new methods are adapted to different situations to enable users in different situations to realize maximum cost savings.
|
