Evaluation of Relative Entropy for Distributed Passive Detection of Weak Acoustic Signals

Variable Scale Relative Entropy Detection for Non-Cooperative Underwater Acoustic Pulse Signals

“…Next, performance comparison between the conventional RE detector [16] and the proposed VSRE detector is made, in terms of the receiver operating characteristic (ROC). The comparison results are shown in Fig.…”

“…8, where the surviving points are divided into two groups. For the group 1, the point with the maximum RE difference has a coordinate of (33, 19), based on which the pulse location is determined as (6.545 : 6.909) s. For the group 2, the point with the maximum RE difference has a coordinate of (19,16), corresponding to the pulse location of (9.474 : 10.105) s. The widths of above detected pulse signals are then estimated as 363.63 ms and 631.58 ms, respectively.…”

“…This verifies the advantage of the proposed VSRE detector. The STFT of packet 8, where the subgraphs represent the pulses detected at (33, 19) and (19,16) by the VSRE detector.…”

“…In the non-cooperative scenario, the goal is achieved by deciding whether the observation sequence has a similar statistical characteristic to an i.i.d probe (ambient noise) sequence, X p ∼ f p (x). To measure the statistical difference between X and X p , the RE has been proposed as a measurement metric [16]. The definition of RE between two PDFs, f p (x) and f (x), is [21]- [23]…”

“…It, however, is unstable without prior information. Motivated by the success of relative entropy (RE) method in multi-input multi-output (MIMO) radar [14], [15], Mignerey studied the feasibility of an RE based distributed passive detection scheme [16]. It is found the resulting RE detector is insensitive to statistical dependence among the sensors and achieves good performance despite of pulse type.…”