Abstract
Detecting and classifying ships based on radiated noise provide practical guidelines for the reduction of underwater noise footprint of shipping. In this paper, the detection and classification are implemented by auditory inspired convolutional neural networks trained from raw underwater acoustic signal. The proposed model includes three parts. The first part is performed by a multi-scale 1D time convolutional layer initialized by auditory filter banks. Signals are decomposed into frequency components by convolution operation. In the second part, the decomposed signals are converted into frequency domain by permute layer and energy pooling layer to form frequency distribution in auditory cortex. Then, 2D frequency convolutional layers are applied to discover spectro-temporal patterns, as well as preserve locality and reduce spectral variations in ship noise. In the third part, the whole model is optimized with an objective function of classification to obtain appropriate auditory filters and feature representations that are correlative with ship categories. The optimization reflects the plasticity of auditory system. Experiments on five ship types and background noise show that the proposed approach achieved an overall classification accuracy of 79.2%, which improved by 6% compared to conventional approaches. Auditory filter banks were adaptive in shape to improve accuracy of classification.
Highlights
Ship radiated noise is one of the main sources of ocean ambient noise, especially in coastal waters.Hydrophones provide real-time acoustic measurement to monitor underwater noise in chosen areas.automatic detection and classification of ship radiated noise signals are still quite difficult at present because of multiple operating conditions of ships and complexity of sound propagation in shallow water
Gammatone filters need to be optimized for the following reasons: (1) There is a fixed bandwidth for a given center frequency. This assumption is not matched by auditory reverse correlation data, which show a range of bandwidths at any given frequency [9]; (2) equivalent rectangular bandwidth (ERB) filter bank cochlea model provides linear filters, which doesn’t account for nonlinear aspects of the auditory system [27]; (3) Auditory filter banks designed from perceptual evidence always focus on the properties of signal description rather than the classification purpose [16]
We proposed an auditory inspired convolutional neural network for ship radiated noise recognition on raw time domain waveform in an end-to-end manner
Summary
Ship radiated noise is one of the main sources of ocean ambient noise, especially in coastal waters. An auditory inspired convolutional neural network is proposed to simulate the processing procedure of human auditory system for ship type classification. The two regions of the auditory system are simulated for ship type classification in a whole model named the auditory inspired convolutional neural network. The proposed model includes three parts: the first part is inspired by the cochlea and takes raw underwater acoustic data as an input This part is performed by a 1D time convolutional layer. A more general way to express the process is: the time convolutional layer yields different simple intrinsic modes of ship noise that help the feature learning at deep layers and help ship targets’ classification at the output layer. The Gammatone filters and features are optimization by CNN to obtain appropriate representations that are correlative with ship categories
Sign-in/Register to view highlights & summary 
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call