Butterfly Optimization-based Normalized Slope Entropy for Ship-radiated Noise Classification

With the in-depth development of sensor technology, human's ability to explore the ocean has been improved. Ship classification and communication can be realized by collecting ship-radiated noise signals (SRNS). However, the complex Marine environment makes these efforts very difficult to achieve. Recently, the feature selection method based on slope entropy has been applied in the field of SRNS classification and has achieved good results. But the two thresholds of slope entropy affect the classification effect. In this study, we developed a normalized slope entropy (NorSE) based on butterfly optimization algorithm (BOA), and proposed BOA-NorSE, which falls the result of slope entropy within the classical interval [0, 1], enhances the divisibility of slope entropy, and uses butterfly optimization to optimize the two thresholds of slope entropy. At the same time, BOA-NorSE is combined with variational mode decomposition and convolutional neural network to form a new classification method for SRNS. Simulation results show that the BOA-NorSE method achieves 81% accuracy in noisy signal recognition. In the measured experiments, compared with other methods based on different entropy, optimization methods and classifiers, the proposed classification method has achieved the highest recognition rate in different data sets in the field of SRNS. This indicates that the proposed method has strong separability and accurate classification accuracy.