arXiv subjects

J. Selva

Publications and source records attributed to J. Selva.

8 recordsLinked to original sources

arXiv2020

Efficient ML Direction of Arrival Estimation assuming Unknown Sensor Noise Powers

This paper presents an efficient method for computing maximum likelihood (ML) direction of arrival (DOA) estimates assuming unknown sensor noise powers. The method combines efficient Alternate Projection (AP) procedures with Newton iterations. The efficiency of the method lies in the fact that all its intermediate steps have low complexity. The main contribution of this paper is the method's last step, in which a concentrated cost function is maximized in both the DOAs and noise powers in a few iterations through a Newton procedure. This step has low complexity because it employs closed-form expressions of the cost function's gradients and Hessians, which are presented in the paper. The method's total computational burden is of just a few mega-flops in typical cases. We present the method for the deterministic and stochastic ML estimators. An analysis of the deterministic ML cost function's gradient reveals an unexpected drawback of its associated estimator: if the noise powers are unknown, then it is either degenerate or inconsistent. The root-mean-square (RMS) error performance and computational burden of the method are assessed numerically.

cs.IT↗

arXiv2017

Wideband Subspace Estimation Through Projection Matrix Approximation

In this paper, we present a wideband subspace estimation method that characterizes the signal subspace through its orthogonal projection matrix at each frequency. Fundamentally, the method models this projection matrix as a function of frequency that can be approximated by a polynomial. It provides two improvements: a reduction in the number of parameters required to represent the signal subspace along a given frequency band and a quality improvement in wideband direction-of-arrival (DOA) estimators such as Incoherent Multiple Signal Classification (IC-MUSIC) and Modified Test of Orthogonality of Projected Subspaces (MTOPS). In rough terms, the method fits a polynomial to a set of projection matrix estimates, obtained at a set of frequencies, and then uses the polynomial as a representation of the signal subspace. The paper includes the derivation of asymptotic bounds for the bias and root-mean-square (RMS) error of the projection matrix estimate and a numerical assessment of the method and its combination with the previous two DOA estimators.

cs.IT↗

arXiv2016

Non-iterative Type 4 and 5 Nonuniform FFT Methods in the One-Dimensional Case

The so-called non-uniform FFT (NFFT) is a family of algorithms for efficiently computing the Fourier transform of finite-length signals, whenever the time or frequency grid is nonuniformly spaced. Following the usual classification, there exist five NFFT types. Types 1 and 2 make it possible to pass from the time to the frequency domain with nonuniform input and output grids respectively. Type 3 allows for both input and output nonuniform grids. Finally, types 4 and 5 are the inverses of types 1 and 2 and are expensive computationally, given that they involve iterative methods. In this paper, we solve this last drawback in the one-dimensional case by presenting non-iterative type 4 and 5 NFFT methods that just involve three NFFTs of types 1 or 2 plus some additional FFTs. The methods are based on exploiting the structure of the Lagrange interpolation formula. The paper includes several numerical examples in which the proposed methods are compared with the Gaussian elimination (GE) and conjugate gradient (CG) methods, both in terms of round-off error and computational burden.

cs.IT↗

arXiv2014

FFT Interpolation from Nonuniform Samples Lying in a Regular Grid

This paper presents a method to interpolate a periodic band-limited signal from its samples lying at nonuniform positions in a regular grid, which is based on the FFT and has the same complexity order as this last algorithm. This kind of interpolation is usually termed "the missing samples problem" in the literature, and there exists a wide variety of iterative and direct methods for its solution. The one presented in this paper is a direct method that exploits the properties of the so-called erasure polynomial, and it provides a significant improvement on the most efficient method in the literature, which seems to be the burst error recovery (BER) technique of Marvasti's et al. The numerical stability and complexity of the method are evaluated numerically and compared with the pseudo-inverse and BER solutions.

cs.IT↗

arXiv2014

Signal Estimation from Nonuniform Samples with RMS Error Bound -- Application to OFDM Channel Estimation

We present a channel spectral estimator for OFDM signals containing pilot carriers, assuming a known delay spread or a bound on this parameter. The estimator is based on modeling the channel's spectrum as a band-limited function, instead of as the discrete Fourier transform of a tapped delay line (TDL). Its main advantage is its immunity to the truncation mismatch in usual TDL models (Gibbs phenomenon). In order to assess the estimator, we compare it with the well-known TDL maximum likelihood (ML) estimator in terms of root-mean-square (RMS) error. The main result is that the proposed estimator improves on the ML estimator significantly, whenever the average spectral sampling rate is above the channel's delay spread. The improvement increases with the spectral oversampling ratio.

cs.IT↗

arXiv2011

Efficient Maximum Likelihood Estimation of a 2-D Complex Sinusoidal Based on Barycentric Interpolation

This paper presents an efficient method to compute the maximum likelihood (ML) estimation of the parameters of a complex 2-D sinusoidal, with the complexity order of the FFT. The method is based on an accurate barycentric formula for interpolating band-limited signals, and on the fact that the ML cost function can be viewed as a signal of this type, if the time and frequency variables are switched. The method consists in first computing the DFT of the data samples, and then locating the maximum of the cost function by means of Newton's algorithm. The fact is that the complexity of the latter step is small and independent of the data size, since it makes use of the barycentric formula for obtaining the values of the cost function and its derivatives. Thus, the total complexity order is that of the FFT. The method is validated in a numerical example.

cs.IT↗

arXiv2010

Efficient Sampling of Band-limited Signals from Sine Wave Crossings

This correspondence presents an efficient method for reconstructing a band-limited signal in the discrete domain from its crossings with a sine wave. The method makes it possible to design A/D converters that only deliver the crossing timings, which are then used to interpolate the input signal at arbitrary instants. Potentially, it may allow for reductions in power consumption and complexity in these converters. The reconstruction in the discrete domain is based on a recently-proposed modification of the Lagrange interpolator, which is readily implementable with linear complexity and efficiently, given that it re-uses known schemes for variable fractional-delay (VFD) filters. As a spin-off, the method allows one to perform spectral analysis from sine wave crossings with the complexity of the FFT. Finally, the results in the correspondence are validated in several numerical examples.

cs.IT↗

arXiv2010

Regularized sampling of multiband signals

This paper presents a regularized sampling method for multiband signals, that makes it possible to approach the Landau limit, while keeping the sensitivity to noise at a low level. The method is based on band-limited windowing, followed by trigonometric approximation in consecutive time intervals. The key point is that the trigonometric approximation "inherits" the multiband property, that is, its coefficients are formed by bursts of non-zero elements corresponding to the multiband components. It is shown that this method can be well combined with the recently proposed synchronous multi-rate sampling (SMRS) scheme, given that the resulting linear system is sparse and formed by ones and zeroes. The proposed method allows one to trade sampling efficiency for noise sensitivity, and is specially well suited for bounded signals with unbounded energy like those in communications, navigation, audio systems, etc. Besides, it is also applicable to finite energy signals and periodic band-limited signals (trigonometric polynomials). The paper includes a subspace method for blindly estimating the support of the multiband signal as well as its components, and the results are validated through several numerical examples.

cs.IT↗