LD approach to asymptotically optimum sensor fusion

摘要: Sensor fusion maybe used to improve detection performance in applications. The idea is make decisions locally, and then transmit them a global centre where the decision made. For making, Bayes or Neyman–Pearson reasoning determines optimal use of local variables. However, determination variables that minimise error probability intractable. In this study, authors design maximise mutual information between binary variable underlying state. This serves as benchmark against which globally optimum solutions compared. Then, they theory large deviations (LDs) determine rule minimises asymptotic probability. LD produces one-dimensional search on receiver operating characteristic curve equalise exponents for false alarm miss probabilities. Many interesting properties solution are proved. Numerical results illustrate asymptotically finite collections sensors.