Blind Array Calibration Technique Using ICA

High-resolution direction of arrival (DOA) estimation algorithms, (e.g. MUSIC algorithms [1]), require array calibration for their mode vectors. One of the calibration techniques is the method which employs a calibration matrix for a covariance matrix of the array. The calibration matrix is estimated before the operation with a set of reference sources having known DOAs in general. Since the method requires known reference signals, it is difficult to apply the method for arrays in actual operation. We propose a array calibration technique with observed signals in the actual operation. Since the observed signals are the mixed signals of several sources, we employ independent component analysis (ICA) [2] [3] to resolve each signal component. The ICA is a method which can resolve mixed signals into individual signal components when the signals are independent and maximally nongaussian. In the proposed calibration technique, the calibration matrix is estimated with these independent signal components. EM (expectation maximization)-type algorithm is employed for the estimation since DOA of the resolved signal components is unknown. In this report, we show the performance of the proposed calibration technique with computer simulations of a 4-element dipole array.