Static magnetic dipole detection using vector linear prediction, Anderson functions, and block-based adaptive processing

We propose a method for combined noise suppression and target detection for 3-axial magnetometer measurements. Vector linear prediction is used to whiten the noise, and matched filters are used for detection. In a case study, we are able to extract a decision variable with an SNR of 20 dB from an input signal where the amplitude of the target is similar to the noise amplitude. A block-based adaptive procedure for use on a buffered stream of input data is also proposed. Good detection performance is achieved on real data, with a detection delay of 85 seconds.