Yuanliang Ma

Moving source localization with a single hydrophone using multipath time delays in the deep ocean

Localizing a source of radial movement at moderate range using a single hydrophone can be achieved in the reliable acoustic path by tracking the time delays between the direct and surface-reflected arrivals (D-SR time delays). The problem is defined as a joint estimation of the depth, initial range, and speed of the source, which are the state parameters for the extended Kalman filter (EKF). The D-SR time delays extracted from the autocorrelation functions are the measurements for the EKF. Experimental results using pseudorandom signals show that accurate localization results are achieved by offline iteration of the EKF.

Localized Multiple Kernel Learning Via Sample-Wise Alternating Optimization

Our objective is to train support vector machines (SVM)-based localized multiple kernel learning (LMKL), using the alternating optimization between the standard SVM solvers with the local combination of base kernels and the sample-specific kernel weights. The advantage of alternating optimization developed from the state-of-the-art MKL is the SVM-tied overall complexity and the simultaneous optimization on both the kernel weights and the classifier. Unfortunately, in LMKL, the sample-specific character makes the updating of kernel weights a difficult quadratic nonconvex problem. In this paper, starting from a new primal-dual equivalence, the canonical objective on which state-of-the-art methods are based is first decomposed into an ensemble of objectives corresponding to each sample, namely, sample-wise objectives. Then, the associated sample-wise alternating optimization method is conducted, in which the localized kernel weights can be independently obtained by solving their exclusive sample-wise objectives, either linear programming (for l1-norm) or with closed-form solutions (for lp-norm). At test time, the learnt kernel weights for the training data are deployed based on the nearest-neighbor rule. Hence, to guarantee their generality among the test part, we introduce the neighborhood information and incorporate it into the empirical loss when deriving the sample-wise objectives. Extensive experiments on four benchmark machine learning datasets and two real-world computer vision datasets demonstrate the effectiveness and efficiency of the proposed algorithm.

Passive localization in the deep ocean based on cross-correlation function matching

Passive localization of a sound source in the deep ocean is investigated in this study. The source can be localized by taking advantage of a cross-correlation function matching technique. When a two-sensor vertical array is used in the deep ocean, two types of side lobe curves appear in the ambiguity surface of the localization. The side lobe curves are analytically expressed and they are then used as indicators of the localization result instead of the scanning point with the maximum power. Simulation and experiment demonstrate the performance of the proposed passive localization method.

A study of the mixed layer of the South China Sea based on the multiple linear regression

Multiple linear regression (MLR) method was applied to quantify the effects of the net heat flux (NHF), the net freshwater flux (NFF) and the wind stress on the mixed layer depth (MLD) of the South China Sea (SCS) based on the simple ocean data assimilation (SODA) dataset. The spatio-temporal distributions of the MLD, the buoyancy flux (combining the NHF and the NFF) and the wind stress of the SCS were presented. Then using an oceanic vertical mixing model, the MLD after a certain time under the same initial conditions but various pairs of boundary conditions (the three factors) was simulated. Applying the MLR method to the results, regression equations which modeling the relationship between the simulated MLD and the three factors were calculated. The equations indicate that when the NHF was negative, it was the primary driver of the mixed layer deepening; and when the NHF was positive, the wind stress played a more important role than that of the NHF while the NFF had the least effect. When the NHF was positive, the relative quantitative effects of the wind stress, the NHF, and the NFF were about 10, 6 and 2. The above conclusions were applied to explaining the spatio-temporal distributions of the MLD in the SCS and thus proved to be valid.

Forward scattering detection of a submerged object by a vertical hydrophone array.

When an object crosses a source-receiver line, the resulting weak acoustic field aberration is often overwhelmed by a strong direct blast. In this study, a lake experiment with a vertical receiver array spanning the water column was designed and conducted with a 10 kHz pulse. The study aimed to detect the aberration caused by forward scattering from an intercepting submerged object. Although such aberration could be directly observed, it varies by only 3 dB, at most, around the direct blast. Hence, the vertical hydrophone array was subjected to time-delay beamforming, and a principal component analysis was conducted on the stable portion of the beam output. The second principal component was extracted from the horizontally directed beam output waveform at the vertical receiver array. The invariant direct blast component was reduced after analysis, and the field aberration caused by forward scattering of the submerged object was amplified by up to 10 dB above the background acoustic field.

Broadband pattern synthesis for circular sensor arrays

The solutions of pattern synthesis are derived for circular sensor arrays based on the criterion of minimizing the mean square error between the desired and synthesized beampatterns. Specifically, the optimal weighting vector, the output beam, and the minimum mean square error are all expressed in closed-form exactly when the desired beampattern is properly formulated. These results provide a more effective and convenient scheme for designing practical frequency-invariant beamformers. Simulations and experimental results demonstrate the performance of the proposed approach.

Range estimation for forward scattering of an underwater object with experimental verification

A range estimation scheme is proposed when a moving object is located within the forward scattering coverage. It is based on acoustic field aberrations of the stable arrival caused by forward scattering of the moving object. The durations of acoustic field aberration are measured by transmitting a signal continuously, and then the range of the moving object is estimated if a priori knowledge of the moving speed is supposed to be known. The acoustic field aberrations of the experimental data are clearly observed after matched filtering, and the estimated ranges of the object agree well with the true values.

Estimating parameter uncertainties in matched field inversion by a neighborhood approximation algorithm

In Bayesian inversion, the solution is characterized by its posterior probability density (PPD). A fast Gibbs sampler (FGS) has been developed to estimate the multi-dimensional integrals of the PPD, which requires solving the forward models many times and leads to intensive computation for multi-frequency or range-dependent inversion cases. This paper presents an alternative approach based on a neighborhood approximation Bayes (NAB) algorithm. For lower dimension geoacoustic inversion, the NAB can approximate the PPD very well. For higher dimensional problems and sensitive parameters, however, the NAB algorithm has difficulty estimating the PPD accurately with limited model samples. According to the preliminary PPD estimation from the NAB, this paper developed a multi-step inversion scheme, which adjusts the parameter search intervals flexibly, in order to improve the approximation accuracy of the NAB and obtain more complete parameter uncertainties. The prominent feature of the NAB is to approximate the PPD by incorporating all models for which the forward problem has been solved into the appraisal stage. Comparison of the FGS and NAB for noisy synthetic benchmark test cases and Mediterranean real data indicates that the NAB provides reasonable estimates of the PPD moments while requiring significantly less computation time.

Matched field noise suppression: A generalized spatial filtering approach

An approach to improving the performance of matched field localization by introducing a generalized spatial filter to suppress interference noise and pass the signal of interest with minimal distortion is presented. The spatial filter is designed by minimizing the maximum distortion of the filtered replica vectors within “passband” while guaranteeing the norm of the filter response within the “stopband” to be lower than some given threshold values. We show that the design problem can be formulated as the second-order cone programming (SOCP) which can be solved efficiently via the well-established interior point method. A modified matched field processor is given to ultimately eliminate the effect of the distortion in the spatially filtered replica vectors. Computer simulation results confirm the effectiveness of the proposed approach by localizing a weak source in the presence of a strong interferer and noise.

Matched field noise suppression: Principle with application to towed hydrophone line array

Discrete noise source suppression in underwater acoustic channel has attracted great attention in recent years. The paper proposes a new principle for dealing with the problem. This new principle is called matched field noise suppression (MFNS). Based on a previous work of the authors group, a full understanding about how a discrete noise source shows effects on the performance of a towed hydrophone line array has been obtained. In light of that finding, MFNS is proposed, which explores and utilizes the characteristics of the noise transmission channel to achieve much greater suppression of the noise in comparison with existing approaches. MFNS combines the concept of matched field processing (MFP) and optimal sensor array processing (OSAP) together to suppress the discrete noise source and to maintain an optimal beam for receiving far-field wanted plane wave signals. A MFNS beam-former is deduced in constraint with signal plane-wave response being unit and noise matched field response being zero. A closed-form solution of the weight vector for the beam-former is given. Computer simulation results agree well to the theoretical analysis.