Search

Performance evaluation of multi-target tracking algorithms is of great practical importance in the design, parameter optimization and comparison of tracking systems. The goal of performance evaluation is to measure the distance between two sets of tracks: the ground truth tracks and the set of estimated tracks. This paper proposes a mathematically rigorous metric for this purpose. The basis of the proposed distance measure is the recently formulated consistent metric for performance evaluation of multi-target filters, referred to as the OSPA metric. Multi-target filters sequentially estimate the number of targets and their position in the state space. The OSPA metric is therefore defined on the space of finite sets of vectors. The distinction between filtering and tracking is that tracking algorithms output tracks and a track represents a labeled temporal sequence of state estimates, associated with the same target. The metric proposed in this paper is therefore defined on the space of finite sets of tracks and incorporates the labeling error. Numerical examples demonstrate that the proposed metric behaves in a manner consistent with our expectations. [ABSTRACT FROM PUBLISHER]

This paper is concerned with recursive estimation using augmented data. We study two recursive procedures closely linked with the well-known expectation and maximization (EM) and space alternating generalized EM (SAGE) algorithms. Unlike iterative methods, the recursive EM and SAGE-inspired algorithms give a quick update on estimates given new data. Under mild conditions, estimates generated by these procedures are strongly consistent and asymptotically normally distributed. These mathematical properties are valid for a broad class of problems. When applied to direction of arrival (DOA) estimation, the recursive EM and SAGE-inspired algorithms lead to a very simple and fast implementation of the maximum-likelihood (ML) method. The most complicated computation in each recursion is inversion of the augmented information matrix. Through data augmentation, this matrix is diagonal and easy to invert. More importantly, there is no search in such recursive procedures. Consequently, the computational time is much less than that associated with existing numerical methods for finding ML estimates. This feature greatly increases the potential of the ML approach in real-lime processing. Numerical experiments show that both algorithms provide good results with low computational cost. [ABSTRACT FROM AUTHOR]

In this correspondence, we focus on blind estimation of the chip duration of time-hopping signals by introducing a cost function based on time-of-arrival (ToA) folding over multiple observation sets. An optimization algorithm that takes advantage of the highly oscillatory behavior in the nearby of the global minimum is proposed and a performance bound for the chip time estimate is derived. The pertinence of our approach is shown through numerical results, considering alternative methods like periodogram and separable least squares line search. The proposed technique enables a good tradeoff between statistical accuracy and computational complexity. [ABSTRACT FROM AUTHOR]

We investigate a linear estimation problem under model uncertainties using a competitive algorithm framework under mean square error (MSE) criteria. Here, the performance of a linear estimator is defined relative to the performance of the linear minimum MSE estimator tuned to the underlying unknown system model. We then find the linear estimator that minimizes this relative performance measure, i.e., the regret, for the worst possible system model. Two definitions of regret are given: first as a difference of MSEs and second as a ratio of MSEs. We demonstrate that finding the linear estimators that minimize these regret definitions can be cast as a Semidefinite Programming (SDP) problem and provide numerical examples. [ABSTRACT FROM AUTHOR]

Systems with quasi-periodically varying coefficients can be tracked using the algorithms known as generalized adaptive notch filters (GANFs). GANF algorithms can be considered an extension, to the system case, of classical adaptive notch filters (ANFs). We show that estimation accuracy of the existing algorithms, as well as their robustness to the choice of design parameters, can be considerably improved by means of compensating estimation delay effects which occur in the amplitude tracking and frequency tracking loops of GANF/ANF filters. Apart from the increased computational burden, the price for the achieved improvements is paid in terms of a decision delay—the proposed generalized adaptive notch smoothing (GANS) algorithms must be run on delayed input/output data sequences. Since such delay is acceptable in many signal processing and system identification applications, the proposed solution seems to be an attractive alternative to the currently available trackers. [ABSTRACT FROM AUTHOR]