Subspace tracking developments for recovery of chirped radar signals

Electronic Surveillance (ES) provides passive surveillance of electromagnetic emissions to achieve detection, location and exploitation of emitters of interest, typically including both communications devices and radar targets. This thesis focuses on development of radar ES capability. The radar signal environment is evolving towards the use of low power wideband modulated emissions, a particular emission type of interest for radar ES being those exhibiting frequency modulation, also known as a 'radar chirp'. Methods such as fixed channelisation of the receiver bandwidth that are typically used in radar ES systems for enhancing system sensitivity and detecting target signals are not optimised to detection of chirped signals, due to the form of the non-stationarity. Research presented in the literature has shown that sensitivity to such emissions can be enhanced by applying blind digital adaptive filtering methods. A particular adaptive approach of interest is that based on projecting noisy input data onto a signal subspace, thus suppressing unwanted noise. This approach can achieve noise suppression in the presence of chirped radar signals but is sub-optimal since a multiple rank solution is required, and the technique is optimal only for statistically stationary signals. This thesis introduces an approach based on transforming the non-stationary input data using a basis function expansion prior to applying subspace-based filtering methods. Three sets of basis functions are investigated: Legendre polynomials, Fourier basis functions, and a set of empirically generated functions. The performance of each in improving performance for block-based SUbspace filtering is investigated. The Fourier functions are also used to develop an extended subspace tracker based on the recursive power method that is also investigated for its performance in filtering linearly chirped signals. SNR gains of up to 10dB in comparison \\lith existing subspace-based methods are demonstrated via these approaches.EThOS - Electronic Theses Online ServiceGBUnited Kingdo