Theoretical and experimental study of a signal feature extraction algorithm for measuring propeller acceleration in a port surveillance system

Classification and identification of moving vessels by extracting particular features of the underwater radiated noise generated by the propellers for use within a port area are an attractive topic for a port underwater surveillance system. The authors first present a chirp-periodic signal model for the envelope of accelerating propeller noise, then derive the maximum-likelihood estimator (MLE) for the evaluation of acceleration and finally propose a novel algorithm for the extraction of acceleration feature on modulated noise (EAFOMN) of the propellers. In addition, a series of experiments were performed for an accelerating propeller in a large cavitation tunnel. The results of the tests show that the actual signal features of accelerating propeller noise is consistent with the presented theoretical analysis, and the proposed algorithm is efficient for passively extracting acceleration of propeller rotation speed.