GRU-NIN model-based underwater acoustic target identification method

Abstract

The invention discloses a GRU-NIN model-based underwater acoustic target identification method, which is based on a multilayer stacked GRU structure, takes implicit states output by all GRUs on the top layer of the multilayer stacked GRU structure as a multi-channel feature map input in MLP convolution operation, and through a layer of 1 * 1 convolution micro-network. integration of multi-channel feature maps obtained by different filters is realized, so that the network learns complex cross-feature map features, and GRU implicit state dimensions are compressed at the same time; the global average pooling layer is connected behind the MLP convolution layer, the spatial average of the feature map after the MLP convolution is calculated, and then the output is fed into the Softmax layer to enhance the corresponding relation between the feature map and the category. Under the framework, the model can realize feature extraction and classification recognition tasks of underwater acoustic target recognition. Experimental results show that the model has better classification and recognition performance than a multi-layer stacked GRU model.

Description

technical field [0001] The invention belongs to the technical field of target recognition, and in particular relates to an underwater acoustic target recognition method. Background technique [0002] Underwater acoustic target recognition is an important part of underwater acoustic signal processing, and it is also an important technical support for underwater acoustic information acquisition and underwater acoustic information countermeasures. Due to the complex marine environment, it is very challenging to use target radiation noise for underwater acoustic target recognition. However, in traditional underwater acoustic target passive recognition systems, feature extraction and classifier are usually two relatively independent links, and the step-by-step processing method does not consider the matching degree of feature extraction and classifier. At the same time, traditional machine learning (ML)-based underwater acoustic target recognition methods mostly use hand-designe...

AI Technical Summary

Problems solved by technology

These hand-designed features rely on expert knowledge and prior knowledge, and their generalization ability is weak; shallow classifiers have limited linear and nonlinear fitting capabilities

Images