A neural network based hybrid system for detection,characterization, and classification of short-duration oceanic signals

A comprehensive classifier system is presented for short-duration oceanic signals obtained from passive sonar, which exhibit variability in both temporal and spectral characteristics even in signals obtained from the same source. Wavelet-based feature extractors are shown to be superior to the more commonly used autoregressive coefficients and power spectral coefficients for describing these signals. A variety of static neural network classifiers are evaluated and are shown to compare favorably with traditional statistical techniques for signal classification. The focus is on those networks that are able to time-out irrelevant input features and are less susceptible to noisy inputs, and two new neural-network-based classifiers are introduced. Methods for combining the outputs of several classifiers to yield a more accurate labeling are proposed and evaluated. These methods lead to higher classification accuracy and provide a mechanism for recognizing deviant signals and false alarms. Performance results are given for signals in the DARPA standard data set I     

Transient sonar signal classification using hidden Markov modelsand neural nets

In ocean surveillance, a number of different types of transient signals are observed. These sonar signals are waveforms in one dimension (1-D). The hidden Markov model (HMM) is well suited to classification of 1-D signals such as speech. In HMM methodology, the signal is divided into a sequence of frames, and each frame is represented by a feature vector. This sequence of feature vectors is then modeled by one HMM. Thus, the HMM methodology is highly suitable for classifying the patterns that are made of concatenated sequences of micro patterns. The sonar transient signals often display an evolutionary pattern over the time scale. Following this intuition, the application of HMM's to sonar transient classification is proposed and discussed in this paper. Toward this goal, three different feature vectors based on an autoregressive (AR) model, Fourier power spectra, and wavelet transforms are considered in our work. In our implementation, one HMM is developed for each class of signals. During testing, the signal to be recognized is matched against all models. The best matched model identifies the signal class. The neural net (NN) classifier has been successfully used previously for sonar transient classification. The same set of features as mentioned above is then used with a multilayer perceptron NN classifier. Some experimental results using “DARPA standard data set I” with HMM and MLP-NN classification schemes are presented. A combined NN/HMM classifier is proposed, and its performance is evaluated with respect to individual classifiers     

A Statistical Analysis of the Detection Performance of a Broadband Splitbeam Passive Sonar

An operational passive sonar is required to detect signals from sources, which are subject to spatial and temporal coherence losses via modifications by the ocean environment. Furthermore, these signals are to be detected in the presence of frequency-dependent correlated noise fields. For a system which employs splitbeam cross-correlation processing, the spatial and spectral properties of the signal and noise are of significant import. Therefore, the exact probability density and cumulative distribution functions of the N-sampled correlator outputs of a splitbeam broadband passive sonar are derived for the case of Gaussian inputs which are described by arbitrary cross-spectral density matrices. The validity of approximating the exact probability density function (pdf) as a Gaussian distribution is investigated. The effect of signal coherence loss and noise correlation on the detection performance is considered and the associated processing loss is expressed as a degradation factor within the detection threshold equation     

The use of spectral information in optimal detection of a source inthe presence of a directional interference

The detection of a passive sonar target in the presence of ambient noise and a plane wave interference is discussed. Intuitively, such a detector consists of a spatial filter which nulls the interference, followed by a temporal filter. In this paper we study the role of the a priori knowledge of the spectrum of the interference and/or signal in improving detector performance. We develop three different generalized likelihood ratio test (GLRT) detectors, resulting from different cases of prior spectral information. We show that, for all cases of known/unknown source and/or interference power spectrum, the GLRT detectors are, as expected, null steering systems. The depth and shape of the null, as well as the postbeamforming temporal filter, are different and are functions of the a priori known spectrum. Under the assumption that all signals and noise are zero-mean Gaussian processes, we analyze the performance of the different detectors and we exploit their dependency on the array beampattern, as well as on the source and interference signal-to-noise ratio. This analysis serves to identify scenarios where the use of prior spectral information leads to significant performance improvement     

基于BP网络对模拟声呐信号分类

针对常规的主动声呐调查设备,在简单海洋分层模型的基础上,模拟了多波束类单频信号、侧扫类单频信号、Ch irp调频信号和混合信号4类声呐接收信号,并针对接收信号特征构造了3层BP网络模型,将隐藏层神经元数目设为可调节;利用时间域脉冲宽度和水深与频率域功率谱密度相结合的特征参量,成功地对模拟信号进行了分类。采用改进的BP网络模型,用训练成功的BP网络对102个检测信号进行了分类测试,结果表明,分类成功率较高,可达76%~84.6%,因而利用BP网络可以对不同类别设备的模拟声呐接收信号进行分类。     

基于频域典型相关分析的水下目标回波识别

水下目标回波的特征提取与分类识别是当前主动声纳关键技术之一。采用基于回波频域特性的典型相关分析算法(CCA:Canonical Correlation Analysis)提取回波的特征,这些特征集中体现了不同目标回波的综合相关特性。设计合适的支持向量机分类器,并获得识别结果。利用这一方法对湖试中的不同目标回波进行分类识别,分析了不同接收信噪比条件下的性能,获得了理想的结果。     

基于小型三维阵列的瞬态声源快速被动定位方法

重物在落水和着底过程中都会产生瞬态声信号,这类信号可被运用于浅水区域水下目标定位。 针对浅水区域目标定位的问题,提出了一种基于小型立体五元基阵的瞬态声源快速被动定位算法。 在分析重物落水信号特征的基础上,选取合适的广义互相关加权函数求得传声器之间的声程差,运用快速最小二乘搜索算法进行声源定位。 结果表明:运用 5 传声器阵列可以同时兼顾定位精度和鲁棒性,且满足实时性要求,该方法可运用于浅水区域瞬态声源定位等领域。     

海底声学底质分类的ELM-AdaBoost方法

基于自适应增强算法（AdaBoost）结合极限学习机（ELM）,通过迭代、调整、优化ELM分类器之间的权值,从而构建了具有强鲁棒性、高精度的ELM-AdaBoost强分类器,增强了现有的ELM分类器的稳定性。以珠江口海区侧扫声呐图像为实验数据,对礁石、砂、泥3类典型底质进行分类识别,该方法的平均分类精度超过90%,优于单一ELM分类器的平均分类精度85.95%,也优于LVQ、BP等传统分类器,且在分类所耗时间上也远少于传统分类器。实验结果表明,本文构建的ELM-AdaBoost方法可有效应用于海底声学底质分类,可满足实时底质分类的需求。     

Humpback whale song or humpback whale sonar? A reply to Au et al

Au et al.'s arguments against the hypothesis that humpback whale songs function as long-range sonar are based on questionable assumptions rather than on empirical data. Like other echolocating mammals (e.g., bats), singing humpback whales: 1) localize targets in the absence of visual information; 2) possess a highly innervated peripheral auditory system; and 3) modulate the temporal and spectral features of their sounds based on environmental conditions. The sonar equation is inadequate for determining whether humpback whale songs generate detectable echoes from other whales because it does not account for temporal variables that can strongly affect the detectability of echoes. In particular, the sonar equation ignores the fact that much of the noise encountered by singing humpback whales is spectrally and temporally predictable, and that audition in mammals is a dynamic and plastic process. Experiments are needed to test the hypothesis that singing humpback whales listen for and respond to echoes generated by their songs     

用于海底底质分类的多波束声强数据选取研究

高质量的海底声强图是进行多波束海底底质分类、目标识别的基础。要得到"单纯"反映海底底质信息的声强图,就需要对原始声强数据进行地形改正,消除地形因素的影响。在描述了多波束数据中水深数据不能满足声强数据的改正要求问题的基础上,提出了以水深数据覆盖范围为约束的声强数据选取方法。实例计算结果表明:该方法在能有效地选取高质量的声强数据,提高了基于声强图像的海底底质分类精度。     

Model-based covariance mean variance classification techniques:algorithm development and application to the acoustic classification ofzooplankton

For inversion problems in which the theoretical relationship between observed data and model parameters is well characterized, a promising approach to the classification problem is the application of techniques that capitalize on the predictive power of class-specific models. Theoretical models have been developed for three zooplankton scattering classes (hard elastic-shelled, e.g., pteropods; fluid-like, e.g., euphausiids; and gas-bearing, e.g., siphonophores), providing a sound basis for model-based classification approaches. The covariance mean variance classification (CMVC) techniques classify broad-band echoes from individual zooplankton based on comparisons of observed echo spectra to model space realizations. Three different CMVC algorithms were developed: the integrated score classifier, the pairwise score classifier, and the Bayesian probability classifier; these classifiers assign observations to a class based on similarities in covariance, mean, and variance while accounting for model spare ambiguity and validity. The CMVC techniques were applied to broad-band (~350-750 kHz) echoes acquired from 24 different zooplankton to invert for scatterer class and properties. All three classification algorithms had a high rate of success with high-quality high SNR data. Accurate acoustic classification of zooplankton species has the potential to significantly improve estimates of zooplankton biomass made from ocean acoustic backscatter measurements     

A comparison of inter-frame feature measures for robust objectclassification in sector scan sonar image sequences

This paper presents an investigation of the robustness of an inter-frame feature measure classifier for underwater sector scan sonar image sequences. In the initial stages the images are of either divers or remotely operated vehicles (ROV's). The inter-frame feature measures are derived from sequences of sonar scans to characterize the behavior of the objects over time. The classifier has been shown to produce error rates of 0%-2% using real nonnoisy images. The investigation looks at the robustness of the classifier with increased noise conditions and changes in the filtering of the images. It also identifies a set of features that are less susceptible to increased noise conditions and changes in the image filters. These features are the mean variance, and the variance of the rate of change in time of the intra-frame feature measures area, perimeter, compactness, maximum dimension and the first and second invariant moments of the objects. It is shown how the performance of the classifier can be improved. Success rates of up to 100% were obtained for a classifier trained under normal noise conditions, signal-to-noise ratio (SNR) around 9.5 dB, and a noisy test sequence of SNR 7.6 dB     

Development and analysis of echo classification using time delays

The results from an investigation of an analytically based method for determining the performance of echo classifiers are presented. In particular, the problem of classifying echo waveforms reflected from objects that are composed of multiple scatterers is considered. The time delays between the multiple echo returns from the individual scattering centers that characterize an object are investigated as features. A generic stochastic point scatterer model is developed for representing the classes of reflecting objects which are of interest. The model allows for uncertainty in prior knowledge about the exact relative location of the individual component scatterers or uncertainty in the delay measurements. A classifier decision algorithm, in the form of a general optimum Bayesian binary classification decision rule suitable for a large variety of classification problems, is derived for the case when the orientation of the reflecting object is known. The case of unknown aspect angle is addressed through the numerical implementation and analysis of two classifiers. The associated performance for all three classifiers is obtained in terms of the probability of error and tied to standard sonar equation parameters. Example binary classification problems are presented and analyzed and some general observations made. A pragmatic framework is established within which complex echo classification issues can be further examined     

Active learning for detection of mine-like objects in side-scan sonar imagery

A data-adaptive algorithm is presented for the selection of the basis functions and training data used in classifier design with application to sensing mine-like targets with a side-scan sonar. Automatic detection of mine-like targets using side-scan sonar imagery is complicated by the variability of the target, clutter, and background signatures. Specifically, the strong dependence of the data on environmental conditions vitiates the assumption that one may perform a priori algorithm training using separate side-scan sonar data collected previously. In this paper, a novel active-learning algorithm is developed based on kernel classifiers with the goal of enhancing detection/classification of mines without requiring an a priori training set. It is assumed that divers and/or unmanned underwater vehicles (UUVs) may be used to determine the binary labels (target/clutter) of a small number of signatures from a given side-scan collection. These sets of signatures and associated labels are then used to train a kernel-based algorithm with which the remaining side-scan signatures are classified. Information-theoretic concepts are used to adaptively construct the form of the kernel classifier and to determine which signatures and associated labels would be most informative in the context of algorithm training. Using measured side-looking sonar data, the authors demonstrate that the number of signatures for which labels are required (via diver/UUV) is often small relative to the total number of potential targets in a given image. This procedure designs the detection/classification algorithm on the observed data itself without requiring a priori training data and also allows adaptation as environmental conditions change.     

Target strength: Some recent theoretical developments

Some recent developments in the theory of diffraction of transient wavefields with high or moderate, but not excessively low, frequency content are reviewed here. Examples are cited to demonstrate that combinations of wavefronts and resonances, which organize the wave process according to progressing and oscillatory events, respectively, can furnish building blocks for interpretation of transient scattering data, and thereby for classification of target features. Conversely, these same spectral elements hold promise as a basis for physically sound parameterization of scattering data to reconstruct target characteristics. The discussion includes impenetrable and penetrable targets, interior-exterior coupling, real and complex spectra, and suggested directions for future research.     

西沙北部海域海洋环境噪声频谱特性

Ambient noise is very important in the prediction system of a sonar performance, because it determines the detection ranges always in a passive sonar and usually in an active sonar. In the uncertainty issue for the so-nar performance, it is necessary to know this factor＇s statistical characteristics that are only obtained by data processing from the underwater ambient noise measurements. Broad-band ambient noise signals from 16 hydrophones were amplified and recorded for 2 min every 1 h. The results show that the ambient noise is essentially depth independent. The cross correlation of the ambient noise levels （1, 6 and 12 h average） with a wind speed is presented. It was found that the correlation is excellent on the upper frequency band and the noise levels correlate better with high wind speed than with low wind speed.     

Sound sources and levels in the ocean

The standard definitions found in the American National Standards on Acoustics are applied to common sound sources used in both underwater acoustics research and naval sonar system operation. Recommended metrics are quantified for both continuous and transient sources of sound. Standard definitions are reviewed with theoretical sound source models. Requisite metrics are derived and applied to examples of energy sources of sound, such as transients from a small omnidirectional explosive, an air gun, a light bulb, and a dolphin click. A generic quantitative model of surface ship sonar system emissions is developed. Active sonar transmissions are analyzed with the requisite quantitative metrics required to characterize these emissions. These results should be useful in environmental assessments, biological experiments, and the sonar system design.     

基于最优特征集的HY-1C卫星海岸带成像仪影像海冰分类方法研究

基于海洋一号C（HY-1C）卫星海岸带成像仪（CZI）遥感影像,提出了一种基于最优特征集的支持向量机海冰分类方法。分别提取CZI影像的光谱特征和纹理特征,采用基于距离可分性的判据进行特征选择,得到最优特征集,以最优特征集作为支持向量机分类器输入,分别对3期辽东湾海域CZI影像开展海冰分类实验和结果分析。结果表明:本文方法得到的海冰分类结果精度优于仅利用光谱特征或纹理特征的海冰分类精度;基于本文方法的3期影像的海冰分类精度均较高,2020年12月19日、2021年1月10日与2021年1月16日的海冰分类总体精度分别为93.67%、91.75%、84.89%,均在80%以上;利用海冰分类结果图估算海冰面积,发现3期辽东湾海冰面积依次增大,最大约为11 998.98 km2。     

Spatial processing of signals received by platform mounted sonar

Acoustic signals received by platform mounted sonar arrays can be spatially processed to enhance the detection of targets in the presence of both ambient and platform generated (self) noise. Ambient noise in the ocean, such as that due to distant shipping or biological choruses, are known to be spatially correlated. The platform generated noise will be of near-field origin and may not be received by all elements in the array. In this paper we investigate the performance of the minimum variance distortionless response (MVDR) beamformer and the recently introduced Fourier integral method (FIM) and compare their performances with the conventional beamformer. Real passive sonar data, obtained from a platform mounted sparse linear array of hydrophones, is used to study the performance of the beamformers in a typical sonar environment. It is shown that in the absence of self noise, when the array is accurately calibrated the MVDR beamformer will perform very well, but when sensor gain or phase errors are present the performance of the MVDR beamformer is degraded. Further, the MVDR beamformer is unable to reject the self noise which is not "seen" by the entire array. FIM however seems to perform well and a modified version of FIM, which we call weighted FIM (WFIM), is shown to perform better and is at worst comparable to a well calibrated MVDR beamformer     

Wavelet analysis of bathymetric sidescan sonar data for the classification of seafloor sediments in Hopvågen Bay - Norway

In this paper we examine the use of bathymetric sidescan sonar for automatic classification of seabed sediments. Bathymetric sidescan sonar, here implemented through a small receiver array, retains the advantage of sidescan in speed through illuminating large swaths, but also enables the data gathered to be located spatially. The spatial location allows the image intensity to be corrected for depth and insonification angle, thus improving the use of the sonar for identifying changes in seafloor sediment. In this paper we investigate automatic tools for seabed recognition, using wavelets to analyse the image of Hopvågen Bay in Norway. We use the back-propagation elimination algorithm to determine the most significant wavelet features for discrimination. We show that the features selected present good agreement with the grab sample results in the survey under study and can be used in a classifier to discriminate between different seabed sediments.