Signal Identification Using a Least L1 Norm Algorithm

In many important applications a signal consists of a sum of exponential terms. The signal is measured at a discrete set of points in time, with possible errors in the measurements. The Signal Identification (SI) problem is to recover the correct exponents and amplitudes from the noisy data. An algorithm (SNTLN) has been developed which can be used to solve the SI problem by minimizing the residual error in the L1 norm. In this paper the convergence of the SNTLN algorithm is shown, and computational results for two different types of signal are presented, one of which is the sum of complex exponentials with complex amplitudes. For comparison, the test problems were also solved by VarPro, which is based on minimizing the L2 norm of the residual error. It is shown that the SNTLN algorithm is very robust in recovering correct values, in spite of some large errors in the measured data and the initial estimates of the exponents. For the test problems solved, the errors in the exponents and amplitudes obtained by SNTLN1 were essentially independent of the largest errors in the measured data, while the corresponding errors in the VarPro solutions were proportional to these largest data errors.

[1]  L. Kaufman.  A variable projection method for solving separable nonlinear least squares problems , 1974 .

[2]  Sabine Van Huffel,et al.  Total least squares problem - computational aspects and analysis , 1991, Frontiers in applied mathematics.

[3]  Åke Björck,et al.  Numerical methods for least square problems , 1996 .

[4]  Lei Zhang,et al.  Accurate Solution to Overdetermined Linear Equations with Errors Using L1 Norm Minimization , 2000, Comput. Optim. Appl..

[5]  S. Kay.  Fundamentals of statistical signal processing: estimation theory , 1993 .

[6]  Mokhtar S. Bazaraa,et al.  Nonlinear Programming: Theory and Algorithms , 1993 .

[7]  Harold W. Sorenson,et al.  Parameter estimation: Principles and problems , 1980 .

[8]  Sabine Van Huffel,et al.  Formulation and solution of structured total least norm problems for parameter estimation , 1996, IEEE Trans. Signal Process..

[9]  Roger Fletcher,et al.  Practical methods of optimization; (2nd ed.) , 1987 .

[10]  R. Fletcher.  Practical Methods of Optimization , 1988 .

[11]  Lennart Ljung,et al.  System Identification: Theory for the User , 1987 .

[12]  Sabine Van Huffel,et al.  Parameter Estimation with Prior Knowledge of Known Signal Poles for the Quantification of NMR Spectroscopy Data in the Time Domain , 1996 .

[13]  Sabine Van Huffel,et al.  The total least squares problem , 1993 .

[14]  Sabine Van Huffel,et al.  Recent advances in total least squares techniques and errors-in-variables modeling , 1997 .

[15]  J. Ben Rosen,et al.  Structured Total Least Norm for Nonlinear Problems , 1998, SIAM J. Matrix Anal. Appl..

[16]  J. Ben Rosen,et al.  Total Least Norm Formulation and Solution for Structured Problems , 1996, SIAM J. Matrix Anal. Appl..

[17]  L. Scharf,et al.  Statistical Signal Processing: Detection, Estimation, and Time Series Analysis , 1991 .

[18]  M. R. Osborne,et al.  Strong uniqueness and second order convergence in nonlinear discrete approximation , 1980 .