Spectral Estimation - LSI has expertise with a number of parametric (e.g., Maximum Entropy or Burg, Modified Covariance, Eigenanalysis or Singular Value Decomposition, etc.) methods of spectral estimation in 
addition to conventional nonparametric (Fourier) processing. Shown here is analysis involved with  defining a baseline signal processing suite for a limited dwell S-band surveillance sensor non-cooperative target identification (NCTI) capability.  In order to overcome the spectral resolution limitations associated with limited coherent dwell operation and conventional processing, autoregressive (AR) non-linear parametric methods of spectral estimation were investigated to provide a relatively modest factor of four improvement in spectral resolution.  The improvement factor was required to enable proven NCTI algorithms to have sufficient resolution for the extraction of discriminating spectral features.  Shown here are sample results of applying the parametric techniques to real target data.  While the spectrum in the upper left corner is the result of conventional processing, the other three results clearly illustrate the ability of parametric techniques to preserve and/or enhance spectral features.  In addition, LSI sourced and advanced other nonconventional nonparametric techniques for coherently combining individual PRF dwells that included up-sampling and down-sampling, fractional time-delay all-pass filter banks and wavelet decomposition-reconstruction filter banks.
 

 
Multisensor, Multitarget Tracking - LSI maintains an avid interest in advancing and applying the latest state estimation concepts to the multisensor, multitarget tracking problem.  LSI’s staff is knowledgeable and experienced in  designing and implementing Kalman and a-b tracking concepts for use in filtering, smoothing (e.g., fixed-point, fixed-lag and fixed-interval) and prediction applications.  The company has a strong background in utilizing data association techniques that range from nearest neighbor to multiple hypothesis testing (MHT) and joint probabilistic data association (JPDA).  Shown here is a steady-state analysis demonstrating the performance penalty (or lack thereof) for track-to-track fusion as compared to measurement fusion.  The plot illustrates the differences in the elements of the error covariance matrix of the fused track for the two approaches, as a function of the measurement accuracy ratio of the two sensors.  The analysis established that less than 7% steady-state performance enhancement is achieved from using measurement fusion as opposed to track-to-track fusion.
 

 
Space Time Adaptive Processing (STAP) - The detection and tracking performance improvements realized by space-time adaptive processing techniques will undoubtedly justify their utilization in next generation  airborne surveillance sensors that possess array antenna architectures. LSI maintains considerable expertise in STAP concepts that range from algorithm development and refinement, to in-situ performance modeling and hardware implementation.  As an illustration of basic STAP concepts, the figure on the left represents a classical STAP architecture with m time taps spaced at the PRI for each of the n spatial channels.  The figures below illustrate the performance of a joint optimum STAP algorithm with an ideal covariance matrix constructed from 16 channels and 32 pulses.  The plot on the lower left indicates an optimized response intended to preserve the indicated look direction and Doppler filter, while simultaneously nulling two jammers and ambiguous platform motion induced antenna sidelobe clutter Doppler.  The plot on the lower right indicates the signal-to-interference plus noise loss that results from applying the STAP algorithm to the entire Doppler space for a particular look direction.  The results indicate that free space performance is recovered to within ~0.25 dB over almost all of the usable Doppler space.