Research On Target Maneuver Identification Model Based On Acoustic Features

Target maneuver detection and tracking technology originated very early and is a very realistic research field.The marine environment is more complex and background noise is widespread.In addition,with the continuous innovation of science and technology,the radiated noise signal of underwater targets is getting smaller and smaller,making the mobile identification of underwater targets has always been a challenging subject.Traditional target maneuver detection is mostly based on the spectral characteristics of the acoustic signal itself.In the marine environment,the target radiated noise often has a low signal-to-noise ratio,and the feature extraction methods based on spectral characteristics cannot achieve good results.This paper takes the radiated noise signal of the underwater target as the research object,establishes the underwater target maneuver recognition model,observes the maneuver state of the underwater target,and calculates its motion parameters during maneuvering to provide the observation platform decision basis.In this paper,proposed a Mel Cepstrum feature extraction method based on empirical mode decomposition.By establishing the underwater target maneuvering identification model,the maneuvering state of the underwater target is observed,and its motion parameters during maneuvering are solved,which provides a decision-making basis for the observation platform.In the process of parameter calculation,a PSO method based on center-random learning,namely CRPSO,is proposed.The center-random learning particle swarm method uses a multipoint optimization method to update the better particles and the center guiding particles and solves the problem that the optimization process caused by the deformed best particles falls into the local optimal solution.It is a kind of center learning and the periodic learning mode combined with random learning enables the optimization method to converge quickly while ensuring the diversity of the calculation process and avoiding premature phenomena as much as possible.In addition to obtaining information and experience through self-exploration,the random distribution-based particle learning method(RPSO)can also effectively use the experience shared by other particles,so that the diversity of each individual in the learning process is improved.Simulation experiments show that the CRPSO method can make full use of the advantages of the central particle learning method's local height search ability and the random learning method's global large-scale random search ability.The application of this method can effectively improve the algorithm's convergence time and identification ability,and it is satisfied result.