Search

Highlights • This paper contributes with two new algorithms within the multiple particle filtering framework that provide an accurate and efficient tool to tackle the challenging problem of filtering in high dimensional state spaces. The proposed algorithms significantly outperform the state-of-the-art in particle filtering for high-dimensional state spaces with low computational burden. • The proposed methods make use of a novel combination of sigma-point integration methods and multiple particle filtering to provide a second order approximation to the marginalization integrals required in the multiple filtering field. • Our approach is not restricted for its use with a specific measurement model and does not require analytic derivations. In addition, the computational cost of the presented algorithms is not increased with respect to similar methods. The resulting particle filters outperform the state-ofthe-art particle filters in high-dimensional problems. Abstract In this paper, we introduce two new particle filtering algorithms for high-dimensional state spaces in the multiple particle filtering approach. In multiple particle filtering, the state space is partitioned and a different particle filter is used for each component of the partition. At each time step, all particle filters share information about their marginal densities so that they can adequately approximate the filtering recursion. In this paper, we propose a second order approximation to the involved densities based on sigma-point integration methods. We then introduce two different particle filters that make use of this strategy. Finally, we demonstrate their remarkable performance through simulations of a multiple target tracking scenario with a sensor network. [ABSTRACT FROM AUTHOR]

Kalman filtering is a widely used framework for Bayesian estimation. The partitioned update Kalman filter applies a Kalman filter update in parts so that the most linear parts of measurements are applied first. In this paper, we generalize partitioned update Kalman filter, which requires the use of the second order extended Kalman filter, so that it can be used with any Kalman filter extension such as the unscented Kalman filter. To do so, we use a Kullback–Leibler divergence approach to measure the nonlinearity of the measurement, which is theoretically more sound than the nonlinearity measure used in the original partitioned update Kalman filter. Results show that the use of the proposed partitioned update filter improves the estimation accuracy. [ABSTRACT FROM AUTHOR]

This paper focuses on the update step of Bayesian nonlinear filtering. We first derive the unscented Gaussian likelihood approximation filter (UGLAF), which provides a Gaussian approximation to the likelihood by applying the unscented transformation to the inverse of the measurement function. The UGLAF approximation is accurate in the cases where the unscented Kalman filter (UKF) is not and the other way round. As a result, we propose the adaptive UGLAF (AUGLAF), which selects the best approximation to the posterior (UKF or UGLAF) based on the Kullback–Leibler divergence. This enables AUGLAF to outperform both the UKF and UGLAF. [ABSTRACT FROM AUTHOR]

Abstract: This paper deals with the problem of tracking using a sensor network when the sensors are not synchronised. We propose a new algorithm called the asynchronous particle filter that, with much less computational burden than the traditional particle filter, has a slightly poorer performance. Thus, it is a good solution to real-time applications with non-synchronised sensors when high performance is required. The low computational burden of the method lies in the fact that we do not predict and update the state every time a measurement is collected. Its high performance is due to the fact that we account for the time instant at which each measurement was taken. [Copyright &y& Elsevier]

In this article, we present the TMM-TPHD and TMM-TCPHD filters, which are the alternative trajectory probability hypothesis density (TPHD) and the alternative trajectory cardinality probability hypothesis density (TCPHD) filters for tracking maneuvering targets. To describe the trajectory maneuver, we define the motion model history sequence of a trajectory as the trajectory motion model (TMM) variable. By incorporating the TMM, we augment the general trajectory model to accommodate the maneuver. Based on this augmented model, we propose the TMM extensions of the standard TPHD and TCPHD filters, namely TMM-TPHD and TMM-TCPHD filters. These filters can directly estimate the number, trajectory state, and TMM of maneuvering targets. In addition to the derivation of the filtering formulas, we also develop linear Gaussian mixture (LGM) implementations for TMM-TPHD and TMM-TCPHD filters. We verify the tracking performance and parameter stability of the proposed filters through tracking simulations involving multiple maneuvering targets. • The TMM variable to express the trajectory motion model history information. • TMM-TPHD/TMM-TCPHD filters are proposed to track maneuvering targets. • Analytic closed-form implementations of proposed filters are developed. • Simulations verify the superior performance in maneuvering target tracking. [ABSTRACT FROM AUTHOR]

• To address the underwater tracking of an unknown and time-varying number of acoustic emitters using passive array sonar systems, a complete particle filter track-before-detect (PF-TBD) signal processing procedure is proposed. • To enhance the detection performance of the low SNR targets, the unthresholded spectrum measurements after the beamforming of the acoustic signals are directly used as the inputs of PF-TBD. • To better model the statistical characteristics of the spectrum measurements, a data fitting based parameter estimation algorithm is proposed to build accurate likelihood functions. • The joint multi-target probability density (JMPD) can be recursively propagated forward by particle filtering to estimate the multi-target states. • To accommodate the time-varying number of targets, the trajectory initiation and termination strategies are also integrated into the filtering process by adaptively adjusting the state dimensions of the JMPD at each measurement time. This work considers the underwater tracking of an unknown and time-varying number of targets, i.e., acoustic emitters, using passive array sonar systems. This problem becomes more challenging if the signal-to-noise ratio (SNR) of the acoustic emitter is low. To address this problem, a complete particle filter track-before-detect (PF-TBD) signal processing procedure is especially developed for the passive array sonar systems. Specifically, in order to enhance the detection performance of the low SNR targets, the unthresholded spectrum measurements after the beamforming of the acoustic signals are directly used as the inputs of the PF-TBD method. To better model the statistical characteristics of the spectrum measurements, a data fitting based parameter estimation algorithm is proposed to build accurate likelihood functions. Then the joint multi-target probability density (JMPD) can be recursively propagated forward by particle filtering to estimate the multi-target states. To accommodate the time-varying number of targets, the trajectory initiation and termination strategies are also integrated into the filtering process by adaptively adjusting the state dimensions of the JMPD at each measurement time. Finally, the efficacy of the proposed PF-TBD method is demonstrated both in simulation and on collected real-world data. [ABSTRACT FROM AUTHOR]

Scribble annotations have recently become popular in video salient object detection. Previous methods only focus on utilizing shallow feature consistency for more integral predictions. However, there is potential for consistency between cross-frame deep features to be used to help regularize saliency predictions better. Besides, we have observed that leveraging saliency predictions as pseudo-supervision signals yields notable improvements in extracting both intra-frame and cross-frame deep features. This, in turn, leads to more precise and detailed object structural information. Thus, we propose a cross-frame feature-saliency mutual reinforcing training process to assist scribble annotations for integral video saliency predictions. Specifically, we design a cross-frame feature regularization head, which leverages intra-frame and cross-frame deep feature consistency to regularize saliency predictions as auxiliary supervision. Then, to help obtain more accurate feature consistency, we design a cross-frame saliency regularization head, where predicted saliency values are used as pseudo-supervision signals to acquire better feature consistency. In this way, our cross-frame feature and saliency regularization heads can benefit from each other to help the network learn more accurately. Extensive experiments show that our method can achieve better performances than the previous best methods. The project is available at https://github.com/muchengxue0911/CFMR. • A mutual reinforcing process is proposed to learn comprehensive object structure. • Intra-/cross-frame feature consistency is deployed to supervise saliency predictions. • Predicted saliency values are used to help build a more accurate feature consistency. [ABSTRACT FROM AUTHOR]

Dynamic Spectrum Access/Cognitive Radio systems access the channel in an opportunistic, non-interfering manner with the primary network. These systems utilize spectrum sensing techniques to sense the occupancy of the primary user. In this paper, an artificial neural network based hybrid spectrum sensing technique is proposed, which considers sensing as a binary classification problem to detect whether the primary user is idle or busy. The proposed scheme utilizes energy detection and likelihood ratio test statistic as features to train the neural network. Moreover, we demonstrate the impact of hyperparameter tuning and carry out the detailed study of it, yielding a combination of best-suited hyperparameters. The performance of the proposed sensing scheme is validated on primary signals of various real world radio technologies acquired with an empirical testbed setup. We conclude that the best performing configuration results in an increase of approximately 63% in detection performance compared to classical energy detection and improved energy detection sensing schemes when averaged over all the radio technologies considered in this work. [ABSTRACT FROM AUTHOR]