The use of d' as a amp;ldquo;decidabilityamp;rdquo; index

摘要

Decision tasks and detection tasks normally involve some uncertainty. The uncertainty is inherent in the task and in the available signal, regardless of strategies that the observer might have, such as being liberal or conservative. The quantity d' is a “criterion-independent” measure which seeks to express how inherently decidable the decision task is, or how detectable the signal is, regardless of the observer's error-avoidance preferences. For the case of biometric identification, d' essentially measures how inherently “discriminable” people are from each other, based on a chosen template and acquisition method. Biometric decisions are fundamentally “same/different” decisions. A same/different decision has four possible outcomes because there are two possible “states of the world” (same or different), and for either of these true situations, there are two possible decisions. Two of these four outcomes are correct. The quantitative statistic d' can be used as a single, dimensionless measure of the inherent decidability of the question “same or different” in the field of biometric identification. When properly derived in accordance with the principles of statistical estimation theory, it can be a useful figure of merit when considering which biometric technology and system is appropriate to a given entry/access control scenario