A robust numerical approach for array calibration

The authors focus on the problem of sensor array calibration in the presence of noise field uncertainties. The objective is to compute a reliable estimate of the array response vector given finite sample estimates of the signal plus noise and noise only array covariances. These covariances are computed from data collected at the sensor array with a calibration signal turned on and off. Numerical difficulties are encountered in applying the usual generalized eigenvalue technique when the noise field is near singular or has changed significantly over the two covariance measurements. A robust numerical approach is presented for determining the response vector in these situations. Results of computer simulations to verify the proposed technique are also presented.<>