A new sequential detector for short-duration signals

A variable threshold page procedure for detection of transient signals

“…1) of all fixed Page tests designed for specific transient signal lengths. From the good match between analysis and simulation results, it appears that the Markov-chain analysis of [7] (synopsized in the Appendix) provides a good estimate of P d when the number of quantization levels is 2 13 .…”

“…set(k = 0) and (Zn = 0) (Z n h(k)) 0! Declare detection: (13) Therefore, the update is g(x) = g 0 (x)0b, where g 0 (1) is a fixed memoryless operation without the bias term. Note that n represents (absolute) sample number, while k refers to the number of samples since the last reset.…”

“…From the experience in [13] it is tempting to choose a light bias, and to tune for performance via the threshold. However, unlike the fixed-style Page test the initial condition on the Page statistic at the beginning of a transient becomes important.…”

“…We propose the VTP procedure since the Page-bank approach needs tuning in the number and location of its Page sub-tests, as the VTP test has performance that can be directly and precisely predicted, and because the VTP test is a single test and hence more efficient. • In [13] a VTP-like heuristic was adopted, where a decision was made to absorb signal-to-noise ratio (SNR) behavior into both threshold and bias: both of these varied with the number of samples since the most recent Page reset, and both were increasing. It turns out that the initially-heavy bias of [13] penalized "long-and-quiet" transient signals too much.…”

“…• In [13] a VTP-like heuristic was adopted, where a decision was made to absorb signal-to-noise ratio (SNR) behavior into both threshold and bias: both of these varied with the number of samples since the most recent Page reset, and both were increasing. It turns out that the initially-heavy bias of [13] penalized "long-and-quiet" transient signals too much. Therefore, the constant-bias VTP choice is preferable.…”