Blind source separation of ship-radiated noise based on generalized Gaussian model

When the distribution of the sources cannot be estimated accurately, the ICA algorithms failed to separate the mixtures blindly. The generalized Gaussian model （GGM） is presented in ICA algorithm since it can model non- Ganssian statistical structure of different source signals easily. By inferring only one parameter, a wide class of statistical distributions can be characterized. By using maximum likelihood （ML） approach and natural gradient descent, the learning rules of blind source separation （BSS） based on GGM are presented. The experiment of the ship-radiated noise demonstrates that the GGM can model the distributions of the ship-radiated noise and sea noise efficiently, and the learning rules based on GGM gives more successful separation results after comparing it with several conventional methods such as high order cumnlants and Gaussian mixture density function.

Blind source separation based on generalized gaussian model

Since in most blind source separation(BSS)algorithms the estimations of probability density function(pdf)of sources are fixed or can only switch between one sup-Gaussian and other sub-Gaussian model,they may not be efficient to separate sources with different distributions.So to solve the problem of pdf mismatch and the separation of hybrid mixture in BSS,the generalized Gaussian model(GGM)is introduced to model the pdf of the sources since it can provide a general structure of univariate distributions.Its great advantage is that only one parameter needs to be determined in modeling the pdf of different sources,so it is less complex than Gaussian mixture model.By using maximum likelihood(ML)approach,the convergence of the proposed algorithm is improved.The computer simulations show that it is more efficient and valid than conventional methods with fixed pdf estimation.

Statistics Modeling of Shallow Sea Ambient Noise and Its Applications in Low-frequency Line Spectrum Detection

The noise's statistical characteristics are very important for signal detection.In this paper,the ambient noise statistical characteristics are investigated by using the recorded noise data in sea trials first,and the results show that the generalized Gaussian distribution is a suitable model for the ambient noise modeling.Thereafter,the optimal detector based on maximum likelihood ratio can be deduced,and the asymptotic detector is also derived under weak signal assumption.The detector's performance is verified by using numerical simulation,and the results showthat the optimal and asymptotic detectors outperform the conventional correlation-integration system due to accuracy modeling of ambient noise.

AN EFFICIENT SHAPE RETRIEVAL WAY

By using the generalized Gaussian density model, this letter puts forward a new shape retrieval way based on the wavelet coefficients. Experimental results show that the proposed shape feature description is superior to the traditional invariant moment algorithm. Moreover, this algorithm provides the important invariant trait to image's size and rotation, which can retrieve images based on shape with more similar results comparing with invariant moment method.

TURBO DECODER USING LOCAL SUBSIDIARY MAXIMUM LIKELIHOOD DECODING IN PRIOR ESTIMATION OF THE EXTRINSIC INFORMATION

A new technique for turbo decoder is proposed by using a local subsidiary maximum likelihood decoding and a probability distributions family for the extrinsic information. The optimal distribution of the extrinsic information is dynamically specified for each component decoder.The simulation results show that the iterative decoder with the new technique outperforms that of the decoder with the traditional Gaussian approach for the extrinsic information under the same conditions.

Mixing matrix estimation of underdetermined blind source separation based on the linear aggregation characteristic of observation signals

Under the underdetermined blind sources separation(UBSS) circumstance,it is difficult to estimate the mixing matrix with high-precision because of unknown sparsity of signals.The mixing matrix estimation is proposed based on linear aggregation degree of signal scatter plot without knowing sparsity,and the linear aggregation degree evaluation of observed signals is presented which obeys generalized Gaussian distribution(GGD).Both the GGD shape parameter and the signals' correlation features affect the observation signals sparsity and further affected the directionality of time-frequency scatter plot.So a new mixing matrix estimation method is proposed for different sparsity degrees,which especially focuses on unclear directionality of scatter plot and weak linear aggregation degree.Firstly,the direction of coefficient scatter plot by time-frequency transform is improved and then the single source coefficients in the case of weak linear clustering is processed finally the improved K-means clustering is applied to achieve the estimation of mixing matrix.The proposed algorithm reduces the requirements of signals sparsity and independence,and the mixing matrix can be estimated with high accuracy.The simulation results show the feasibility and effectiveness of the algorithm.

A New Approach for the DFT NIST Test Applicable for Non-Stationary Input Sequences

The National Institute of Standards and Technology (NIST) document is a list of fifteen tests for estimating the probability of signal randomness degree. Test number six in the NIST document is the Discrete Fourier Transform (DFT) test suitable for stationary incoming sequences. But, for cases where the input sequence is not stationary, the DFT test provides inaccurate results. For these cases, test number seven and eight (the Non-overlapping Template Matching Test and the Overlapping Template Matching Test) of the NIST document were designed to classify those non-stationary sequences. But, even with test number seven and eight of the NIST document, the results are not always accurate. Thus, the NIST test does not give a proper answer for the non-stationary input sequence case. In this paper, we offer a new algorithm or test, which may replace the NIST tests number six, seven and eight. The proposed test is applicable also for non-stationary sequences and supplies more accurate results than the existing tests (NIST tests number six, seven and eight), for non-stationary sequences. The new proposed test is based on the Wigner function and on the Generalized Gaussian Distribution (GGD). In addition, this new proposed algorithm alarms and indicates on suspicious places of cyclic sections in the tested sequence. Thus, it gives us the option to repair or to remove the suspicious places of cyclic sections (this part is beyond the scope of this paper), so that after that, the repaired or the shortened sequence (original sequence with removed sections) will result as a sequence with high probability of random degree.

L1/2 Regularization Based on Bayesian Empirical Likelihood

Bayesian empirical likelihood is a semiparametric method that combines parametric priors and nonparametric likelihoods, that is, replacing the parametric likelihood function in Bayes theorem with a nonparametric empirical likelihood function, which can be used without assuming the distribution of the data. It can effectively avoid the problems caused by the wrong setting of the model. In the variable selection based on Bayesian empirical likelihood, the penalty term is introduced into the model in the form of parameter prior. In this paper, we propose a novel variable selection method, L1/2 regularization based on Bayesian empirical likelihood. The L1/2 penalty is introduced into the model through a scale mixture of uniform representation of generalized Gaussian prior, and the posterior distribution is then sampled using MCMC method. Simulations demonstrate that the proposed method can have better predictive ability when the error violates the zero-mean normality assumption of the standard parameter model, and can perform variable selection.

Improved Particle Filter for Non-Gaussian Forecasting-aided State Estimation

Gaussian assumptions of non-Gaussian noises hinder the improvement of state estimation accuracy.In this paper,an asymmetric generalized Gaussian distribution(AGGD),as a unified representation of various unimodal distributions,is applied to formulate the non-Gaussian forecasting-aided state estimation problem.To address the problem,an improved particle filter is proposed,which integrates a near-optimal AGGD proposal function and an AGGD sampling method into the typical particle filter.The AGGD proposal function can approximate the target distribution of state variables to greatly alleviate particle degeneracy and promote precise estimation,through considering both state transitions and latest measurements.For rapid particle generation from the AGGD proposal function,an efficient inverse cumulative distribution function(CDF)sampling method is employed based on the derived approximation of inverse CDF of AGGD.Numerical simulations are carried out on a modified balanced IEEE 123-bus test system.The results validate that the proposed method outperforms other popular state estimation methods in terms of accuracy and robustness,whether in Gaussian,non-Gaussian,or abnormal measurement errors.

Speckle Reduction Based on Contourlet Transform Using Scale Adaptive Threshold for Medical Ultrasound Image

A new speckle suppression method in contourlet domain was presented. By modeling the subband contourlet coefficients of the ultrasound images after logarithmic transform as generalized Gaussian distribution (GGD), we gave a scale-adaptive threshold in Bayesian framework. Experimental results of both synthetic and clinical ultrasound images show that our method has a better performance on speckle suppressing than the wavelet-based method while well preserving the feature details.

Effective and Robust Detection of Adversarial Examples via Benford-Fourier Coefficients

Adversarial example has been well known as a serious threat to deep neural networks(DNNs).In this work,we study the detection of adversarial examples based on the assumption that the output and internal responses of one DNN model for both adversarial and benign examples follow the generalized Gaussian distribution(GGD)but with different parameters(i.e.,shape factor,mean,and variance).GGD is a general distribution family that covers many popular distributions(e.g.,Laplacian,Gaussian,or uniform).Therefore,it is more likely to approximate the intrinsic distributions of internal responses than any specific distribution.Besides,since the shape factor is more robust to different databases rather than the other two parameters,we propose to construct discriminative features via the shape factor for adversarial detection,employing the magnitude of Benford-Fourier(MBF)coefficients,which can be easily estimated using responses.Finally,a support vector machine is trained as an adversarial detector leveraging the MBF features.Extensive experiments in terms of image classification demonstrate that the proposed detector is much more effective and robust in detecting adversarial examples of different crafting methods and sources compared to state-of-the-art adversarial detection methods.

Identification of important factors influencing nonlinear counting systems

Identifying factors that exert more influence on system output from data is one of the most challenging tasks in science and engineering.In this work,a sensitivity analysis of the generalized Gaussian process regression(SA-GGPR)model is proposed to identify important factors of the nonlinear counting system.In SA-GGPR,the GGPR model with Poisson likelihood is adopted to describe the nonlinear counting system.The GGPR model with Poisson likelihood inherits the merits of nonparametric kernel learning and Poisson distribution,and can handle complex nonlinear counting systems.Nevertheless,understanding the relationships between model inputs and output in the GGPR model with Poisson likelihood is not readily accessible due to its nonparametric and kernel structure.SA-GGPR addresses this issue by providing a quantitative assessment of how different inputs affect the system output.The application results on a simulated nonlinear counting system and a real steel casting-rolling process have demonstrated that the proposed SA-GGPR method outperforms several state-of-the-art methods in identification accuracy.

Design and implementation of the NaI(Tl)/CsI(Na) detectors output signal generator

We designed and implemented a signal generator that can simulate the output of the NaI（Tl）/CsI（Na）detectors＇pre-amplifier onboard the Hard X-ray Modulation Telescope（HXMT）.Using the development of the FPGA（Field Programmable Gate Array）with VHDL language and adding a random constituent,we have finally produced the double exponential random pulse signal generator.The statistical distribution of the signal amplitude is programmable.The occurrence time intervals of the adjacent signals contain negative exponential distribution statistically.