Detection of Gauss-Markov Random Field on Nearest-Neighbor Graph

The problem of hypothesis testing against independence for a Gauss-Markov random field (GMRF) with nearest-neighbor dependency graph is analyzed. The sensors measuring samples from the signal field are placed IID according to the uniform distribution. The asymptotic performance of Neyman-Pearson detection is characterized through the large-deviation theory. An expression for the error exponent is derived using a special law of large numbers for graph functionals. The exponent is analyzed for different values of the variance ratio and correlation. It is found that a more correlated GMRF has a higher exponent (improved detection performance) at low values of the variance ratio, whereas the opposite is true at high values of the ratio.