Change-Point Detection Of Weak Signals: How To Use Signal Correlation?

We explore the problem of how to exploit known signal temporal correlation is detecting weak signal. We model the signal as a time series with a known signal correlation structure, and proposed a novel maximum score test (MST) for weak signal detection. The MST avoids the computational expensive inversion of covariance matrix in the maximum likelihood test. We develop analytic approximation of significant level and standardized power function of MST, which are shown to be very accurate by simulation. We compare the MST detector with the maximum likelihood detector without using signal correlation.