Research And Application Of Target Classification And Recognition Algorithm Based On Underwater Acoustic Signal

With the development of marine resources and detection technology,it is necessary to conduct in-depth research on the classification and recognition of acoustic signal targets.The classification and recognition of underwater acoustic signals aims to improve the precision of underwater target recognition,there are two main research directions: one is to seek effective feature extraction methods,and the other is to design more effective classification and recognition algorithms.Target classification recognition is based on acoustic signal data,using preprocessing,feature extraction and classification recognition methods for fast and accurate determination of target identity.To perform effective feature extraction,design a more efficient classification and recognition system and achieve efficient processing of the data,the main research work is as follows:(1)Improvements to feature extraction algorithms.For the feature extraction by MFCC is not rich enough,the improved feature extraction by MFCC algorithm is proposed by combining EEMD and MFCC feature extraction.The pre-emphasis coefficients and decomposition have been adjusted to replace complex phase multiplication operations with simple summation and shift operations and to obtain a more detailed division of the signal,resulting in a more detailed division of the signal.(2)Improvements to the quantum particle swarm algorithm.Due to the complex non-linear behavior of particle motion,an adaptive quantum particle swarm algorithm based on exponential functions is proposed.A nonlinear function is used to describe the dynamic change law of CE coefficient during the iterative process.The value of changing index is equivalent to the convexity and concavity of the curve of changing CE coefficient.Studying the performance of AQPSO optimization is equivalent to studying the convexity and concavity of nonlinear changing CE coefficient curve.(3)Improvements to the classifier.Due to the better application of SVM on acoustic signals,the AQPSO algorithm accelerate speeds up the convergence speed of the algorithm.This also increases the diversity of particles when iterating and reduces the probability of premature maturation.The study proposes the use of the AQPSO algorithm to train SVM classifiers,resulting in a new classifier.