Constrained generalized likelihood ratio test in distributed detection

In distributed detection system based on local statistics, the performance of traditional sum fusion will degrade if the difference among the local signal-to-noise ratios (SNRs) is large. Based on theoretical analysis of this degradation, we propose to replace conventional MLE with constrained maximum likelihood estimation (MLE) and corresponding constrained generalized likelihood ratio test (GLRT). Result shows that the constrained GLRT can provide improvement on detection probability over that of sum fusion rule.