Olivier Rabaste

Estimation of Multipath Channels With Long Impulse Response at Low SNR via an MCMC Method

This paper addresses the estimation of multipath channels with long impulse response at low signal-to-noise ratio (SNR). The channel sparseness impulse response is modeled by means of a Bernoulli-Gaussian process. Then, the optimization of the resulting posterior distribution resorts to Monte Carlo Markov chain (MCMC) simulation techniques. Special attention is devoted to noise correlation involved by matched filtering: Taking this correlation into account in the algorithm derivation leads to significantly improved performance for both amplitude and time-delay estimation. The method is also extended to cope with Doppler frequency offsets. In particular, simultaneous paths with different Doppler offsets can be estimated. Crameacuter-Rao lower bounds (CRLBs) are derived and presented together with simulation results

Rejection of Doppler shifted multipaths in airborne passive radar

In this article, we consider the problem of rejecting multipath ground echoes in airborne passive radar. In such an application, these multipath echoes present some Doppler-shift due to the movement of the airborne platform. Classic rejection methods with rejection mask sampled at the cell resolution in delay and Doppler lead to strong sidelobe residue that tend to mask the target responses. We propose here first to oversample this mask, solving the projection problem induced using the Moore-Penrose pseudo-inverse. This solution however requires the knowledge of the right oversampling factor necessary to get good rejection performance. We thus propose a different approach that permits to perform rejection over a continuous set, thus insuring good rejection. This solution provides some insight on the dimensionality of the subspace spanned by this continuous set, which can in turn be used to properly choose the oversampling factor. Simulations are provided for rejection on a Doppler interval, and in the context of airborne passive radar dealing with range and Doppler.

Mismatched filter optimization for radar applications using quadratically constrained quadratic programs

This paper deals with the problem of optimal filter computation for specific radar applications.We consider here a method for finding the mismatched filter that minimizes the peak-to-sidelobe ratio. This method is based on a reformulation of the optimization problem as a convex quadratically constrained quadratic program (QCQP) and ensures that any found solution is a global solution of the problem. This formulation also enables the insertion of additional constraints, e.g., a constraint on the loss in processing gain of the optimal filter. We study in this context the robustness of the proposed filter to phase errors. Then, we propose to adapt this QCQP to deal with Doppler shifts and provide results for specific applications. Next, we investigate the possibility to use mismatched filters to reduce the high sidelobe level observed along the range axis of multiple-input multiple-output (MIMO) ambiguity functions produced by phase codes. We propose two MIMO receiver architectures that enable the integration of mismatched filters. The first architecture tries to minimize all autocorrelations and cross-correlations of the transmitted phase codes, while the second architecture computes mismatched filters adapted to the signals transmitted in different angular direction. The latter approach enables us to obtain a MIMO ambiguity function close to the perfect thumbtack shape.

A particle filter for target arrival detection and tracking in Track-Before-Detect

In this paper, we address the problem of detecting the appearance time of a target and tracking its state with a particle filter in the Track-Before-Detect context. We show that it is possible to model the problem as a quickest detection change problem in a Bayesian framework. In this case, the posterior density of the target time appearance is a mixture where each component represents the hypothesis that the target arrived at a given time. As the posterior density is intractable in practice, we propose to approximate each component of the mixture by a particle filter, and we show that the weights of the mixture can be computed recursively thanks to quantities provided by the different particle filters. The overall filter yields good performance.

An MCMC algorithm for BOC and AltBOC signaling acquisition in multipath environments

The binary offset carrier (BOC) and alternate BOC (AltBOC) waveforms, introduced for Galileo signaling, are designed to enable high time delay estimation accuracy. However, their autocorrelation functions have multiple narrow peaks, thus leading to potential acquisition bias, in particular in presence of multipaths. In order to detect the first path, a high sampling rate is considered. Unfortunately, this introduces noise correlation. In this paper, we compare a technique developed for multipath channel estimation based on Monte Carlo Markov chain (MCMC) simulation with the standard acquisition approach, in multipath environments, considering both accuracy and computational load issues.

Matching pursuit via continuous resolution cell rejection in presence of unresolved radar targets

In this article a new matching pursuit algorithm with continuous radar resolution cell rejection is proposed. It allows matching pursuit to work well even if more than one target is present in some resolution cell (unresolved targets) of the radar matched filter: it prevents its tendency to generate spurious sidelobes or miss a weaker target hidden in stronger target sidelobes. The FMCW radar case is particularly investigated which offers a very natural and computationally inexpensive solution to the problem that can also be applied in spectral analysis. The extension of the proposed approach to any radar waveform is also investigated.

Around-the-corner radar: Detection and localization of a target in non-line of sight

This paper examines the problem of detecting and locating an NLOS target in an urban environment by exploiting multipath radar returns. We propose a detection-localization algorithm based on a matched subspace filter approach that works in the target state space directly. A real experimentation was carried out to show that a portable radar can provide images of multipath returns in NLOS cases that can be clearly interpreted thanks to a simple ray tracing model. The application of the detection-localization algorithm on experimental radar measurements shows that the target can be detected and located but that the mitigation of strong ambiguities inherent to the multipath detection-localization problem remains a challenging problem for the application at hand.

Multitarget likelihood computation for track-before-detect applications with amplitude fluctuations of type swerling 0, 1, and 3

Track-before-detect methods jointly detect and track one or several targets from raw sensor measurements. They often require the computation of the measurement likelihood conditional on the hidden state, which depends on the complex amplitudes of the targets. Since these amplitudes are unknown and fluctuate over time, this likelihood must be marginalized over the complex amplitude (i.e., phase and modulus). It has been demonstrated that this marginalization can be done analytically over the phase in the monotarget case. In this article, we first propose to extend the marginalization to the modulus in a monotarget setting, and we show that closed forms can be obtained for fluctuations of type Swerling 1 and 3. Second, we demonstrate that, in a multitarget setting, a closed form can be obtained for the Swerling 1 case. For Swerling 0 and 3 models, we propose some approximation to alleviate the computation. Since many articles consider the case of squared modulus measurements, we also consider this specific case in monoand multitarget settings with Swerling 0, 1, and 3 fluctuations. Finally, we compare the performance in estimation and detection for the different cases studied, and we show the gain, both in detection and estimation, of the complex measurement method over the squared modulus method, for any fluctuation model.

Improving BEM channel estimation for airborne passive radar reference signal reconstruction

In this article, we consider an airborne passive radar using ODFM signals of opportunity in a Single Frequency Network (SFN) environment. The long and sparse multipath fading channel and the mobility of the receiver can deeply alterate the signal and therefore degrade the decoding processing. Classic methods experience difficulties in dealing with such channel impacts. Consequently we present a Basis Expansion Model (BEM) based channel estimation method. We also propose to combine it with the Minimum Description Length (MDL) and ESPRIT algorithms to get a prior information. Simulation results emphasize the benefits of using the BEM and confirm the necessity to have an a priori knowledge of the channel.

Geometrical design of radar detectors in moderately impulsive noise

The presence of crossing seas in the nearshore maylead to a drastic amplification of local wave heights or to a substantial change in the orientation of the highest parts of the wave crest owing to nonlinear interactions of waves in shallow water. The location and strength of the related effects can be roughly forecast based on the properties of crossing wave systems in the framework of the Kadomtsev-Petviashvili equation. We introduce a method of the identification of crossing seas and singling out the properties of interacting wave systems from numerically simulated two-dimensional wave energy spectra of selected locations in the Baltic Sea obtained within a multi-decadal (1956-1997) wave hindcast using the WAM model. Each spectrum spans across 24 evenly spaced directions and 40 frequencies starting from 0.042 Hz (23.8 s) to about 1.718 Hz (0.58 s). The numerically replicated spectra usually contain a certain level of noise, which may lead to the detection of false maxima and is filtered out using a Gaussian-type convolution filter. We then test each sample of the resulting anti-aliased distribution with a pyramid shaped stencil in order to find the spectral density, frequency and direction of all relative maxima. Their frequency and direction is then mapped back onto the initial spectra to evaluate the heights of the single wave systems. The average reduction of maxima detection from unfiltered to filtered data is 9.2%.