Cyclostationarity for Passive Underwater Detection of Propellor Craft: A Development of DEMON Processing

As the blades of a propeller pass through the water they produce characteristic amplitude modulated random noise signals which can be detected using sonar. A popular empirical technique for passive acoustic detection of surface ships from submarines using these sonar signals is DEMON (Detection of Envelope Modulation on Noise) processing. As the name suggests, DEMON processing seeks to detect the frequencies of modulation, i.e. the shaft and blade pass frequencies. It works by isolating the frequency band in which the modulation is most distinct to the operator, taking the envelope of this filtered noise band and producing a waterfall spectrogram. Harmonics associated with the rotating components of the propeller will be manifest in the waterfall, allowing the vessel to be identified. DEMON processing has several drawbacks however, most importantly the requirement for operator skill in the selection of the noise band. This paper presents the preliminary findings of work underway to provide a mathematical formalisation of the empirical DEMON processing technique. This formalisation is based on the observation that DEMON processing unknowingly exploits the cyclostationary properties of the propeller signals. Cyclostationary signal processing, a technique which has recently found application in mechanical systems involving rotating machinery, offers great insight into the detection problem, and has the potential to overcome the weaknesses of DEMON processing as well as expanding its capability to include frequency as well as amplitude modulation.