Information theory based sensor surveillance

Abstract

Hydrophones phase and gain dispersions have a deep impact on conventional beampatterns of line arrays, by affecting the sidelobe level. Indeed, high sidelobe levels threaten the detection of weak sources in the presence of strong jammers. Sensors failures are even more critical. Sensors surveillance algorithms are therefore essential to the array performances. They often consist in selecting valid sensors whose power spectral densities are close to a certain estimated mean within a certain fixed or estimated standard deviation. These statistics estimations first take the assumption of no sensors failures, and require parameters settings. After having recalled the impact of sensors dispersions and sensors failures on conventional beampatterns, parameter free sensors surveillance algorithms are proposed. They are based on information criteria, such as Stochastic Complexity Minimization or Akaike Information Criteria. These sensors selection methods are compared to the more traditional methods described above on synthetic data and sea trial signals.