Interference detection in Gaussian noise

Abstract

Interference detection in Gaussian noise is proposed. It can be applied for easy detection and editing of interference lines in radio spectral line observations. One does not need to know the position of occurrence or keep track of interference in the band. By using statistical properties of N-channel Gaussian noise it is possible to differentiate interference from normal Gaussian noise. These statistical properties are three quantities which are calculated on the subject spectrum. The first is the expected absolute maximum from the mean level across the spectrum. The second is the expected absolute difference maximum (array of adjacent differences of channel values taken across the spectrum), and the third is the expected absolute added difference maximum. For N-channel Gaussian noise, these quantities have a well-defined value. Any deviation across the spectrum that violates an upper limit formed by these expected maxima is attributed to interference. Results obtained on real data by applying this technique have been displayed.