Search

This paper shows the recognition of syllables through the using of Wavelet Transform and statistical tools such as cross covariance and the Pearson correlation coefficient. The considered syllables in this work are Consonant-Vowel (CV) and Vowel-Consonant (VC) given that they constitute 60% of those present in the Spanish language. The signals of the syllables are processed through a focused analysis with the application of time windows (segmentation the syllables in their respective phonemes) to later perform a multi-resolution analysis using the Wavelet Transform. For the recognition of the CV or VC structure, the cross covariance and the Pearson correlation coefficient are used. The results shows the performance of proposed method in comparison with other methods.

This paper considers the distributed fusion problem where estimates in local processors are transmitted to a fusion center. An improved matrix-weighted fusion estimation framework is proposed for the systems with one-step delays from local processors to fusion center. In this framework, the one-step predictions instead of local estimates are transmitted to the fusion center. In fusion center, the one-step predictions are weighted by a matrix to obtain the final estimates. In order to use matrix-weighted algorithm, a recursive form for the cross covariance matrices of one-step prediction errors between any two local processors are derived. A simulation for the system with three sensors verifies that the one-step prediction in fusion center has higher precision than that in each local processor, and our method has smaller error than the general matrix-weighted fusion estimation algorithm when there exists one-step delays from local processors to fusion center.

This paper demonstrates a signal processing approach designed to overlay data sets measured using two optical micrometers. Two sensors are used in order to measure object's shape, but measured signals are relatively shifted, because it is impossible to mount sensors into the same position. Covariance method are applied to measurement data and the shift between these sets are evaluated.

The practice of inertial sensor calibration has commonly been carried out by taking into account the deterministic and stochastic components of the error measurements issued from a calibration session. Once the deterministic components have been taken into account through physical models, the remaining stochastic component has always been dealt with for each sensor separately. The latter process involves estimating complex probabilistic models for each sensor which has been proven to be extremely complicated over the past years, although recent proposals have allowed to overcome most of the limitations that have characterized this task. However, the separate stochastic calibration of the individual sensors composing an inertial measurement unit may not be wise in many cases since there can be considerable degrees of dependence between the sensors, especially between the gyroscopes. For this reason, there has been growing attention towards this issue in order to consider the influence of the stochastic behaviour of the sensors on each other, with few proposals that address this problem. Among these proposals there has been the idea of integrating the information coming from the different gyroscopes so as to build a virtual gyroscope. In this paper we build on this idea and, using a recently proposed method for multivariate signal modelling, we deliver a general and flexible framework that allows to consider many different modelling options which provide the basis to construct a virtual sensor that optimally combines the information from the individual sensors and considerably improves navigation accuracy.

The atomic norm minimization (ANM) has been successfully incorporated into the two-dimensional (2-D) direction-of-arrival (DOA) estimation problem for super-resolution. However, its computational workload might be unaffordable when the number of snapshots is large. In this paper, we propose two gridless methods for 2-D DOA estimation with L-shaped array based on the atomic norm to improve the computational efficiency. Firstly, by exploiting the cross-covariance matrix an ANM-based model has been proposed. We then prove that this model can be efficiently solved as a semi-definite programming (SDP). Secondly, a modified model has been presented to improve the estimation accuracy. It is shown that our proposed methods can be applied to both uniform and sparse L-shaped arrays and do not require any knowledge of the number of sources. Furthermore, since our methods greatly reduce the model size as compared to the conventional ANM method, and thus are much more efficient. Simulations results are provided to demonstrate the advantage of our methods.

Foreground detection in complex scenarios is a challenging task. In this work, we propose to detect foreground by incrementally learning a cross-covariance based subspace. In our method, we first introduce the cross-covariance based two dimensional principal component analysis (2DPCA) algorithm into foreground detection field for better background estimation. Then, we extend the conventional cross-covariance based 2DPCA algorithm into an incremental one, which helps to model background in an adaptive way. Moreover, we consider the sparse and the continuous characteristics of the foreground, and formulate them as a fused lasso problem. By adding the fused lasso regularization into the proposed subspace learning process, we integrate the background recovery and the foreground estimation into a single optimization framework. Finally, we design an efficient approximate algorithm which solves the optimization problem effectively. We compare our method with the state of the art methods on multiple challenging video sequences. The experimental results demonstrate the effectiveness and the advantages of the proposed method.

In this paper, a novel 2D direction of arrival (DOA) estimation approach based on tensor technique for the conformal array is proposed. In order to improve the estimation performance with low SNR and small snapshots, a third-order measurement tensor is constructed by exploiting the multidimensional information of the conformal array received data. Then, the desired signal subspace is obtained by the higher-order singular value decomposition (HOSVD) of the cross covariance tensor. Finally, the DOAs could be estimated by the conventional subspace based algorithms. This approach enables us to get a better DOA estimation performance owing to utilize the multidimensional information and the cross covariance tensor method, especially in the case of low SNR and small number of snapshots. At last, the simulation results demonstrate the validity and the advantages of the proposed method.

In this paper, two dimensional angle estimation using a new array structure named separate nested acoustic vector sensor (AVS) array is studied. Unlike traditional AVS array, where the pressure sensor and velocity sensors are collocated in space, the separate nested AVS array has two separate and nested subarrays including the pressure sensor array and the velocity vector sensor array. With the extended vectorization of the cross covariance matrix between the two arrays, 2M2-1 degrees of freedom (DOF) can be achieved with only M physical acoustic vector sensors. After analyzing the fitting error, only one dimensional sparse representation is required to estimate the two dimensional angle with automatically pairing. Compared to conventional AVS array based methods, the proposed scheme can achieve better angle estimation performance and handle more sources. Simulation results verify the effectiveness of the proposed method.

24th Signal Processing and Communication Application Conference, SIU 2016; Zonguldak; Turkey; 16 May 2016 through 19 May 2016, Bu çalışmada, elektrotlar arası zaman gecikmesi kullanılarak bir beyin-bilgisayar ara yüzü çalışması gerçekleştirilmiştir. Öznitelik olarak, seçilen referans kanalı ile geriye kalan tüm kanalların çapraz kovaryansının mutlak değerinin en yüksek olduğu zaman gecikmeleri hesaplanmıştır. Çalışmada kullanılan 5 kişi içinden 3 kişinin sınıflandırma performansının %100’e yakın olmasının yanında bu kişilerin eğitim veri seti sayısının diğer iki kişiye göre oldukça düşük olması ve literatürde buna benzer çalışmaların azlığı, önerilen yaklaşımın geliştirilmeye açık olduğunu göstermektedir., In this study, we have carried out a brain-computer interface study that uses time delays between electrodes. As features, we have calculated the time delays that maximizes the absolute value of the cross-covariance between the chosen reference channel and the remaining channels. Performance results of 3 out of 5 participants that were nearly at a %100 accuracy level along with a relatively smaller number of training data, and a lack of similar studies indicate that the proposed approach is open to the further improvements.

DNA-binding proteins play a pivotal role in various intra- and extra-cellular activities ranging from DNA replication to gene expression control. With the rapid development of next generation of sequencing technique, the number of protein sequences are unprecedentedly increasing. Thus it is necessary to develop computational methods to identify the DNA-binding protein from the protein sequence information. In this study, a novel method is presented which combines the support vector machine and the auto-cross covariance transformation. The protein sequence represented in the form of amino acids or the physical-chemical properties of amino acids are converted into a series of fixed-length vectors by Kmer composition and the auto-cross covariance transformation. The sequence order effect can be effectively capture by this scheme. These vectors are then inputted to support vector machine to discriminate the DNA-binding proteins from the non DNA-binding ones. The proposed method achieves the overall accuracy of 75.23% and Matthew correlation coefficient of 0.5 by a rigorous jackknife test. The independent test shows that the proposed method outperforms most of the existing methods. These results demonstrate that the proposed method provides the state-of-the-art performance for the prediction of DNA-binding proteins.

For cellular communication systems, CCI (Co-Channel Interference) and ACI (Adjacent Channel Interference) are becoming a capacity limitation. IRC(Interference Rejection Combining) is a technique used in an antenna diversity or virtual antenna diversity system to suppress CCI and ACI, by using the cross covariance between the noise in diversity channels or multi samples. Under interference scene, the IRC technique can obtain many gain compared to the convention receiver. However, under sensitivity scene, the IRC technique worsens the performance compared to the conventional receiver. In order to overcome this drawback of IRC technique, a scene detection algorithm is proposed. The simulation results show this scene detection algorithm, proposed in this paper, can keep the IRC performance in the interference scene and the conventional receiver performance in the sensitivity scene simultaneously

It is often hard to settle the estimation problems for the signal systems with time delays. By modern time series analysis method, the systems with time delays can be transformed into those without time delays. By the measurement predictor and the white noise estimators, the local and the optimal information fusion Wiener signal estimators are presented. Applying the CI (Covariance Intersection) method, the CI fused Wiener signal estimators are derived, which avoids the calculation of the cross covariance matrx between local sensors. Their estimation accuracy is higher than those of the local Wiener estimators. A Monte-Carlo simulation result shows that the actual accuracy of the presented CI fusion Wiener smoother approximates to that of the corresponding optimal information fusion smoother, and based on the covariance ellipse, the geometric interpretation of the accuracy relation is shown.

The distributed optimal fusion problem for the state estimation of multi-sensor discrete-time stochastic systems with additional state equality constraints is considered. After obtaining the filtering error cross covariance between local filters, by the distributed optimal fusion algorithm weighted by matrices, the constrained distributed optimal fusion filter is derived. A numerical example shows that the fusion filter is better than each local filter.

For a target tracking problem, such as tracking of a mobile robot or an unmanned vehicle, multiple sensors are required to achieve accurate estimated position of the target. Practically, measurements from sensors arrive out of sequence, e.g., delayed data due to the processing of images. We call these measurements Out of Sequence Measurements (OOSMs). Many researches propose solutions to OOSMs using an Extended Kalman filter (EKF) or particle filter (PF) as a basic algorithm. Our previous research proposes an algorithm that applies Ensemble Kalman filter (EnKF) to handle the OOSM problem. We store ensembles of the state particles during the filtering process and make use of the information about those ensembles later. By calculating a cross covariance between ensembles from different points of time, we can directly update the current estimated state with delayed measurements. Moreover, by using EnKF, we can simply apply the method to systems with strong nonlinear models without finding any Jacobian or backward transition matrix. However, our previous algorithm only preforms well for one-step lag measurements. In order to handle multistep lag measurement, in this paper, we propose an algorithm with an additional backward updating step. We illustrate the results of simulations comparing with Rauch-Tung-Striebel (RTS) smoothing filter and the conventional algorithms proposed by [1] and [2] which apply EKF and particle filter techniques, respectively. The proposed algorithm shows commendable results compared to others.

A two dimensional (2-D) DOA estimation method is proposed for non-circular sources in this paper. According to the non-circularity of the sources, we utilize both the receiving signal and its conjugate to estimate the DOA angles aiming at expanding the array virtually, which can improve the estimation properties at low signal to noise ratio (SNR) and small number of array elements. Then based on the principle of ESPRIT, the DOA angles are estimated. The proposed new method utilizes the properties of the receiving cross covariance matrix and simple pair matching method to cut down the computation load. Finally, a 2-D DOA estimation method for coherent sources is proposed via spatial smoothing techniques. Simulation results show that the proposed algorithm can estimate independent and coherent sources effectively and has much better performances than that of common ESPRIT algorithms.

Kalman Filter is a very important tool in multi-sensor data fusion. One problem with the Kalman Filter is that it requires either that the measurements are independent or that the cross-covariance or correlation is known. However, cross-covariance among measurements from different local sensors is inevitable owning to common process noise and not easily calculated owning to insufficient information and high computational complexity. So a recent research emphasis focuses on seeking new methods of fusing state vectors and their covariance or simplifying current methods. In this paper, a new geometric fusion method called covariance coverage is proposed, which not only needn't to consider cross-covariance between measurements, but also has low computational complexity. What's more, covariance coverage method has strong extensibility and can directly support the fusion of tracking system who has more than two local sensors. Simulation experiments and results show that the accurateness of fusion state estimate by covariance coverage method is clearly higher than that of each local sensor and a little letter than that of covariance intersection algorithm proposed by J. Julier.

Direction-of-arrival (DOA) estimation for multiple sources using the array of three-component electromagnetic (EM) vector-sensors is considered in this paper within a tensorial framework. The cross covariance tensor is firstly defined as the one in which the cross covariance terms between every two components of the array signals are arranged in a tensorial manner, and then the trilinear decomposition is performed to this tensor for the parallel factors, from which DOA information for the impinging signals could be extracted. The proposed method does not require any a priori knowledge on the array configuration and is shown to offer high robustness to colored noise. Simulations are provided to illustrate the performance of the proposed DOA estimator.

Normally cokriging, which is extended by kriging to take into account extra information and viewed as an unbiased prediction method with minimum prediction variance, integrates the information carried by a secondary variable related to the primary variable being estimated. The linear model of coregionalization (LMC) and the Markov model 1 (MM1) are proposed for cokriging to fulfill the integration of the primary variable and the secondary one. The main limitation of LMC is the requirement of modeling a positive definite cross covariance matrix for both primary and secondary variables, which cannot be solved by original cokriging. Although MM1 is a reasonable model if the primary variable is defined on the same or a larger volume support than the secondary one, allowing it to screen the influence of further away data of the primary variable; consider the case of a secondary variable defined on a much larger support than the primary variable, the MM1 is not appropriate. Then an improved Markov model 2 (MM2) for such a case is presented to meet the above condition. The MM2 screening hypothesis indicates that the secondary datum screens the influence of all further away secondary data on primary datum. Experimental results show that cokriging under the MM2 is practical when a secondary variable is defined on a much larger support than the primary variable.

A practical method is proposed to segment the wrist pulse waveform and estimate the average waveform. Some key issues that would affect the performance of the tasks are addressed. A zero-phase filtering was used to accommodate low frequency variations and high frequency noise without the phase-shift distortion, and a moving-window adaptive threshold based segmentation algorithm was used to ensure the segmenting performance. Waveform rotating and scaling, outlier elimination, cross-covariance based alignment, and average waveform estimation were introduced. Testing results show the effectiveness of segmentation performance, and the resulting average waveform well reflect the typical characteristics of the analyzed wrist pulse trend.

By the modern time series analysis methods, based on ARMA innovation model and augmented state space model, a multisensor optimal information fusion Wiener deconvolusion predictor weighted by scalars is proposed. The formulas of computing the local predictor error variances and cross-covariances are given, which are applied to compute optimal weighting coefficients. Compared to the single sensor case, the accuracy of the fused predictor is improved. A simulation example shows its effectiveness.

Based on the optimal fusion algorithm weighted by scalars in the linear minimum variance sense, a distributed optimal fusion reduced-order Kalman filter with scalar weights is presented for discrete-time stochastic singular systems with multiple sensors and correlated noises. It has higher accuracy than any local filter does. Compared with the distributed fusion filter weighted by matrices, it has lower accuracy but has reduced computational burden. Computation formula of cross-covariance matrix of the filtering errors between any two sensors is given. An example with three sensors shows the effectiveness.

This paper describes techniques for track level fusion of surveillance data that are applicable to existing and near term tactical surveillance systems. The linear optimal fused estimate is a convex combination of remote estimates with weights being the estimation confidences (covariances). The covariance based algorithm is most applicable where the track estimate is generated by a Kalman filter based tracking system. When track covariance is not available, such as in /spl alpha/-/spl beta/ tracking systems, an estimated covariance can be used for track fusion. In addition, track fusion also requires accounting for the cross covariance between tracks. Various approaches to estimating the auto covariances and the cross covariances are examined, and the performance is evaluated through computer simulations.

The necessary and sufficient conditions for the almost sure convergence of adaptive filtering algorithms based on (a/n) stepsize are established. Assuming that the covariance of the output signal (Y/sub n/) and the cross-covariance of (Y/sub n/) and the reference signal ( psi /sub n/) are constant, if (Y/sub n/ Y/sub n/') satisfies a law of large numbers (where ' denotes transpose), then the necessary and sufficient condition for almost sure convergence of adaptive filtering algorithms is that (Y/sub n/ psi /sub n/') also satisfies the law of large numbers. Moreover when there exists a small deviation from the law of large numbers for (Y/sub n/ Y/sub n/') and (Y/sub n/ psi /sub n/'), there is also a bounded deviation dependent on the former from the convergence of the algorithms, and the latter tends to zero as the former goes to zero. >

This paper focuses on the extension of the asymptotic covariance of the sample covariance (denoted Bartlett's formula) of linear processes to third-order sample cumulant and to noisy linear processes. Thanks to a matrix polyspectral representation, closed-form expressions of the asymptotic covariance and cross covariance of the sample second and third moments are derived in a straightforward manner. As an application of these extended formulae, we enhance the sensitivity of the asymptotic performance of estimated ARMA parameters by an arbitrary third order-based algorithm to the spectrum of colored additive noise.

Considers the effects on track association of two features that are often neglected in analysis. These two features are the cross-covariance between two tracks of the same target and the effects of a loss of resolution in a sensor. The sensors modelled in this paper are nonhomogeneous in both the state space and the measurement space, and a method for calculating the cross-covariance for such dissimilar sensors is derived. Also, a simple resolution model is proposed. The paper uses Monte Carlo simulations to illustrate the effects of these two features on a number of track association methods.