| In complex underwater environments,the data received by sonar typically includes one or more radiated acoustic source signals.The mutual interference of multiple radiated acoustic source signals causes the collected radiated noise signal from underwater targets to become a multi-component signal,which also contains noise and interference.The multi-component signal hinders the extraction of signal features,thereby limiting the performance of underwater target recognition.To address this issue,this paper combines the intrinsic mode functions with time-series deep learning,and based on experimental data,constructs an underwater target recognition model that combines three different decomposition algorithms(empirical mode decomposition,ensemble empirical mode decomposition,and complementary ensemble empirical mode decomposition)with time-series deep learning.The model constructed in this paper is evaluated using model evaluation metrics.This paper first investigates the basic concepts of empirical mode decomposition,ensemble empirical mode decomposition,and complementary ensemble empirical mode decomposition,and applies these three decomposition methods to the radiated noise signals from underwater targets.All three decomposition algorithms yield a series of intrinsic mode functions.The correlation coefficients between the IMFs obtained by the three algorithms and the original signals are calculated using the correlation coefficient method.The results indicate that the IMFs obtained by CEEMD exhibit the strongest correlation with the original signals.Next,the IMFs based on EMD,EEMD and CEEMD,along with time-series deep learning networks,are combined to construct an underwater target recognition model.A comparative analysis is performed to assess the impact of different learning rates and optimization algorithms on the recognition results of underwater targets.Additionally,the effects of different decomposition algorithms on the recognition results are compared.Finally,the proposed approach of combining intrinsic mode functions with time-series deep learning is applied to process lake experimental data,validating the feasibility of this method.Furthermore,it is confirmed that the recognition results of the underwater target recognition method based on CEEMD and timeseries deep learning outperform those based on EMD and EEMD in terms of the intrinsic mode functions and time-series deep learning. |
