| System Identification Toolbox | |
Segmentation of Data
Sometimes the system or signal exhibits abrupt changes during the time when the data is collected. It may be important in certain applications to find the time instants when the changes occur and to develop models for the different segments during which the system does not change. This is the segmentation problem. Fault detection in systems and detection of trend breaks in time series can serve as two examples of typical problems.
The System Identification Toolbox offers the function segment to deal with the segmentation problem. The basic syntax is
with a format like rarx or rarmax. The matrix thm contains the piecewise constant models in the same format as for the algorithms described earlier in this section.
The algorithm that is implemented in segment is based on a model description like (3-58), where the change term 
is zero most of the time, but now and then it abruptly changes the system parameters 
. Several Kalman filters that estimate these parameters are run in parallel, each of them corresponding to a particular assumption about when the system actually changed. The relative reliability of these assumed system behaviors is constantly judged, and unlikely hypotheses are replaced by new ones. Optional arguments allow the specification of the measurement noise variance 
in (3-57), of the probability of a jump, of the number of parallel models in use, and also of the guaranteed lifespan of each hypothesis. See segment in the "Command Reference" chapter.
| | Available Algorithms | Some Special Topics | |