Time-optimal recognition method with the multi-level distance function

The multi-level distance function recognition algorithm consists in modifying the list of classes to which the recognized object can belong by eliminating the wrong classes with the use of the distance-type decision criteria. The distance functions used in the criteria form a set which can be treated as a single, multi-level function. The list initially contains all the known classes. The decision criteria, i.e. the levels, are used sequentially. The recognition time depends on the time of checking each criterion and the list lengths on the levels. The latter depend on the order of levels. Therefore, a fixed time-optimal sequence can be found for a given set of classes. If in this set the surplus functions are included, it is possible to find the optimal subset to be used in the recognition system. The expected list lengths are estimated with the use of probabilities of erroneous class assignments made by decision criteria