An Analysis of Automatic Testing Threshold Criteria

This paper develops a method of selecting a decision threshold when the uncertainties in the levels of the signal to be measured and the threshold detector are Gaussian. The false alarm probability and undetected fault probability are established as a function of mean threshold level and the parameters of the uncertainties. These two errors are combined, and this combination is plotted for various weightings of the two types of errors. In general, optimum threshold settings exist. The use of multiple thresholds for reducing the probability of making the wrong decision is also discussed.