Adaptive ship-radiated noise recognition with learnable fine-grained wavelet transform

Analyzing the ocean acoustic environment is a tricky task. Background noise and variable channel transmission make it complicated to implement accurate ship-radiated recognition. Existing recognition systems are weak in addressing underwater environment, thus leading disappointing performance practical application. In order keep system robust various environments, this work proposes an adaptive generalized — AGNet (Adaptive Generalized Network). By converting fixed wavelet parameters into fine-grained learnable parameters, learns characteristics of sound at different frequencies. Its flexible design conducive capturing more background information (e.g., noise, channel). To utilize implicit spectrograms, adopts convolutional neural network with parallel convolution attention modules as classifier. Experiments reveal that our outperforms all baseline methods on several datasets, could benefit from transfer learning. Moreover, shows against interference factors.