Automatic filter design and calibration

All analog devices introduce phase distortion to the signals that pass through them. As of today, most digital signal processing systems that operate on these signals do not correct for the phase distortion introduced by analog hardware since it is remarkably difficult to use standard filter design algorithms to automatically and reliably construct compensating digital filters. Consequently, hardware designers work hard to minimize the distortion caused by their devices, effectively sacrificing performance elsewhere. 
TNG has developed a new technology for robust automatic design of accurate and efficient filters which, in particular, can be applied to phase compensation filters. The entire design process is performed by robust and non-iterative algorithms so that the resulting filter achieves the user-selected accuracy and can be embedded into any device without supervision from an expert. 

Rapid resampling of time series to or from arbitrary grids

A ubiquitous and critically important problem in signal processing is the digital resampling of signals. For example, resampling is needed for signals distorted by Doppler effects due to emitter and receiver motion. Although the problem of digital resampling has a long history and it is straightforward to convert between two sample rates that are related by a simple rational factor, existing algorithms make it very challenging to resample an input signal at an arbitrary time-varying sampling rate. Existing solutions to this problem generally use a polyphase filter bank to produce an approximate solution, but this approach sacrifices accuracy and bandwidth in order to achieve a reasonable computational cost. TNG has developed and implemented a new fast algorithm that completely solves this problem yielding high accuracy and preserving almost the entire bandwidth, with a computational cost that is superior to traditional polyphase solutions. 

Highly efficient interpolation on irregular grids with large missing regions

Spatial data is usually collected on irregular grids which are often undersampled in some locations. For example, due to orbital constraints, satellite data are generally collected on irregular grids which contain large missing regions, or cloud cover may prevent obtaining information of different sections of the target area; dealing with such data is a major processing challenge for both defense and scientific missions. We have developed a multistage recursive algorithm that constructs a functional representation of this type of data by adaptively refining the representation. At each stage, we match the band limit of the interpolating function to the mesh granularity, yielding a stable algorithm. The resulting interpolating function is available in a highly efficient functional form, so it can be rapidly and accurately evaluated at arbitrary output locations. TNG is successfully using this new algorithm for sea surface temperature data processing and is currently developing additional novel algorithms to perform a variety of tasks for that type of data.