Source separation based processing for integrated Hall sensor arrays

Integrated arrays of Hall-type magnetic sensors are generally subject to significant crosstalk due to poor spatial selectivity of very closely spaced sensors. In this paper, we explore blind source separation as a signal processing technique to unmix a number of magnetic sources that impinge on the array. The processing consists of two stages: the first one estimates the source number and projects the observations on the signal subspace; the second stage is a source separation algorithm. Experimental results from a silicon array of Hall sensors, interfaced with a DSP, demonstrate real-time separation of two magnetic sources even under ill-conditioning of the mixing matrix.