Adaptive algorithms for solving generalized eigenvalue signal enhancement problems
Introduction
In many signal processing applications, we are presented with a time sequence of composite M-dimensional vectorswhere is a linear manifold of some underlying desired signal s(n) and is undesired noise or interference. If the signals are Gaussian, maximum likelihood estimators of the signal employ linear processing to maximize the output signal power while minimizing the output interference power. Here we assume that this is the case and form an instantaneous linear estimate of s(n) aswhere is a “weight” vector that forms the linear estimate and superscript H denotes (Hermitian) complex conjugate; complex signals are assumed throughout for generality.
Examples of such a setup abound in the literature, where the signal and interference vectors could represent samples of a random process in the temporal domain [4] or array element signals in the spatial domain [3]. The temporal domain formulation arises in classical signal detection and estimation problems, while spatial-domain scenarios involve array processing of electromagnetic or acoustic signals.
We express the output signal and interference power as, respectively,andwhere are the respective covariance matrices associated with the signal and interference components of . The intention of the adaptive algorithm is to adjust so as to make the output signal power “large” and the output interference power “small”. One objective function to realize these two goals can be formulated as the ratioHowever, maximization of this ratio is often not satisfactory for a number of reasons. First of all, we need to form estimates of the signal covariance matrices, so that maximization of (5) may lead to a poorly conditioned solution that is susceptible to additive noise and other unmodeled effects. Also, in some applications, it may be desirable to “balance” the degree to which Ps is maximized and Pu is minimized. An example of this is in downlink cellular mobile communications, where is the weighting of an antenna array and the problem is to control how much emphasis should be placed on maximizing the power to the intended mobile, Ps, as compared to the resources spent on minimizing power to the co-channel mobiles, Pu [8]. In both cases, the problem is solved by “regularizing” the denominator of (5) to obtain the modified objective functionwhere α is a parameter whose value determines the balance of resources expended in maximizing Ps on one hand and minimizing Pu on the other. If α=0, then ξ=ξ0 and we have the unconstrained version (5), which typically favors minimization of Pu. At the other extreme, if α→∞, then Ps is maximized without any regard to Pu. Clearly, some finite value of α will result in the best performance, as determined by the particular application.
Objective functions like , are known as Rayleigh quotients and their maximization is well known to be the solution of a generalized eigenvalue problem involving the numerator and denominator matrix pair. Thus, the solution to maximizing (6) is the eigenvector corresponding to the largest eigenvalue of the generalized eigenvalue problemHowever, we do not know the exact values of the signal and interference covariance matrices, and , and so must estimate these from the available data sequence . Here, we assume that the signal and interference components can be somehow isolated to provide individual training samples of signal vectors and interference vectors . This can be accomplished, for example, if we have side information as to when either alone or alone are present in . Another way this can be provided is for example in a communications environment, whereby a known training sequence of data symbols enables the signal vector to be perfectly reconstructed from the detected message data symbols and subtracted from the composite received signal to obtain the interference signal vector. Following this rationale, we take and as given quantities.
In this paper we use the signal and interference vectors to obtain adaptive solutions of the generalized eigenvalue problem (7). The most straightforward approach is to use these vectors to form estimates of the associated covariance matrices, which are then substituted into (7) to set up the generalized eigenvalue problem. This two-step process is the most straightforward to understand. However, it is computationally demanding, especially when the vector dimension M is large. As a first simplification we will propose a hybrid technique that simultaneously updates the covariance matrix estimates and computes an iterative solution to the optimal generalized eigenvector. To further improve numerical efficiency, we will also derive stochastic gradient algorithms that recursively update the optimal solution of in one step for each new set of data vector snapshots. In the next section, we discuss these algorithms. Then in Section 3, we introduce a simple wideband signal model and evaluate the performance of the various algorithms in Section 4. Conclusions are presented in Section 5.
Section snippets
Optimal (OPT)
A benchmark is first established for the best performance that could be achieved when the signal and interference covariance matrices and are known exactly. The generalized eigenvalues and eigenvectors corresponding to the pair {} can therefore be readily calculated using any number of standard routines [1, p. 375], e.g., as embodied in the Matlab eig command.
Sample matrix/generalized eigenvector (SMGEV)
For real applications, we need to estimate the covariance matrices from vector snapshots. The usual way to do this is to
Signal model
In this section, we define the signal model used in the next section to evaluate the adaptive algorithms of Section 2. For concreteness, this model will be formulated in the spatial domain for a linear antenna array with multiple fading paths. However, similar models could be easily envisaged using alternative interpretations in the temporal domain, such as finite-bandwidth signals that experience independent random attenuation of spectral components. Further commentary will follow later after
Methodology
In this section, we will evaluate the performance of the various algorithms of Section 2 using the spatial signal model of Section 3. The performance is quantified by the convergence of the signal power and interference power as a function of the number of independent data vectors n. (Recall from Section 3 that for simplicity we have chosen unit signal and noise powers at the element level.) As representative angle spreads, we choose two values Δs=Δu=3° and 30°, corresponding
Conclusions
In this paper, we have developed and evaluated several adaptive algorithms for solving generalized eigenvalue signal enhancement problems using signal and interference vector samples. Direct two-step approaches provide the highest performance but are at best O(M2) algorithms, where M is the dimensionality of the vectors. A stochastic gradient algorithm reduces the complexity to O(M) but convergence is very sluggish, especially when the signal and interference overlap. However, an O(M)
References (13)
- et al.
Matrix Computations
(1996) Adaptive Filter Theory
(1991)- et al.
Array Signal Processing: Concepts and Techniques
(1993) Fundamentals of Statistical Signal Processing: Detection Theory
(1998)Method of stochastic approximation in the determination of the largest eigenvalue of the mathematical expectation of random matrices
Automat. Remote Control
(1970)- et al.
A quasi-Newton adaptive algorithm for generalized symmetric eigen-value problem
IEEE Trans. Signal Process.
(1996)
Cited by (26)
Robust adaptive modified newton algorithm for generalized eigendecomposition and its application
2007, Eurasip Journal on Advances in Signal ProcessingOnline Dominant Generalized Eigenvectors Extraction via a Randomized Algorithm
2023, IEEE Transactions on Vehicular TechnologyOn the Discrete-Time Dynamics of Cross-Coupled Hebbian Algorithm
2022, Journal of Computer Science and TechnologyOnline dominant generalized eigenvectors extraction via a randomized method
2021, European Signal Processing ConferenceUnified and self-stabilized parallel algorithm for multiple generalized eigenpairs extraction
2020, IEEE Transactions on Signal Processing