Marine Vessels Acoustic Radiated Noise Classification in Passive Sonar Using Probabilistic Neural Network and Spectral Features

Underwater target classification at greater depths using deep neural network with joint multiple‐domain feature

“…In [2], the wavelet transform (WT) is used to extract tonal features from the average power spectral density (PSD) while two neural network classifiers are used to evaluate the classification results. In [3], a new configuration of probabilistic neural network (PNN) called multi‐spread PNN (MSPNN) is proposed for marine vessel classification with the features of autoregressive (AR) model. In [4, 5], different feature extraction approaches using wavelet packet demonstration (WPT) are developed.…”

An automated approach to passive sonar classification using binary image features

“…The proposed classifier is a kind of back propagation (BP) neural network (Duda et al, 2000) with some modifications. Two sets of discriminating features were proposed in (Farrokhrooz and Karimi, 2011). The first set of features is extracted from AR model of radiated noise and the other is directly extracted from power spectral density of radiated noise.…”

SONAR Systems and Underwater Signal Processing: Classic and Modern Approaches

“…In this section, we evaluate the classification performance of conventional PNN and MSPNN with the method in (Farrokhrooz 2011) for spread estimation by applying them to the bank of real acoustic radiated noises of marine vessels. We define the performance of a classifier for a class as the probability of correct decision for the data of that class.…”

“…Another disadvantage of the conventional PNN is using the same spread parameter for all classes, which decreases the degree of freedom of the PDF estimator. To overcome these disadvantages, a multi-spread PNN that uses a different spread parameter for each class has been recently introduced (Farrokhrooz 2011). Consider an exemplar classification problem with two discriminating features that is illustrated in figure 9.…”

“…The proposed classifier is a kind of Back Propagation (BP) neural network (Duda 2000) with some modifications. Two sets of discriminating features proposed in (Farrokhrooz 2011). The first set of features is extracted from AR model of radiated noise and the other is directly extracted from power spectral density of radiated noise.…”