Detection of radio-frequency interference in microwave radiometers using spectral kurtosis

This paper describes the spectral kurtosis detector as an additional indicator for radio frequency interference, RFI in passive remote sensing systems. The estimator is based on continuous Fast Fourier Transformation of samples, followed by evaluation of each frequency bin in subsequent data blocks, using a kurtosis estimator. The spectral kurtosis response is simulated and tested in a laboratory setup for various RFI types, and the estimator response is evaluated and compared to the response of the traditional time domain kurtosis estimator. Results show that great benefits of spectral kurtosis is reached for continuous wave RFI or for RFI with high duty-cycle compared to the radiometer integration time. Typically the spectral kurtosis is superior for duty-cycles above 15%, while standard kurtosis is more efficient for lower duty-cycles, down to a few percent. In combination, the two estimators provide a very good indicator for most RFI types.