An optimal decision rule has been derived by Chair and Varshney (1986) for fusing decisions based on the Bayesian criterion. However, to implement such a rule, the miss probability PM and the probability of false alarm PF for each local detector must be known, and these are not readily available in practice. To circumvent this situation, an adaptive fusion system for equiprobable sources has been developed. The system is extended to unequiprobable sources; thus its practicality is enhanced. An adaptive fusion model using the fusion result as a supervisor to estimate the PM and Pp is introduced. The fusion results are classified as 'reliable' and 'unreliable'. Reliable results are used as a reference to update the weights in the fusion centre. Unreliable results are discarded. The convergence and error analysis of the system are demonstrated theoretically and by simulations. The paper concludes with simulation results that conform to the analysis.
All Science Journal Classification (ASJC) codes
- Computer Networks and Communications
- Distributed detection
- Probability of false alarm
- Reinforcement learning