Signal Fusion-based Distributed Detection in Heterogeneous Radar Scenarios

In distributed radar system, local radar may form heterogeneous statistics which probability density functions (PDFs) are different, but it is typically difficult to combine such observations for better detection performance. In this paper, taking fusion of the cell-average constant false alarm rate detector and the square-law detector for instance, we propose a signal-fusion based distributed detection algorithm and its global test statistic is a sum of linearly weighted local statistics, which the local weights depend on the channel SNRs. According to the Neyman-Pearson criterion and under the assumption of independence of local observations, the optimal test statistic is a sum of complicated functions of local test statistics and thus its PDFs are difficult to obtain. We first find the equivalent local test statistics to ensure that the PDF of a sum of the equivalent local test statistics under the null hypothesis has a closed-form expression. Then we approximate the global test statistic by the first-order Taylor series of the local likelihood ratio functions and local test statistics are properly weighted. Numerical results show that the detection performance of the proposed method outperforms the decision fusion algorithm. It is also found that the proposed algorithm is robust to SNR estimation errors.