In recent years, Prof. Levy has examined robust estimation, filtering, equalization and detection problems. The objective of this research is to develop signal processing techniques which are insensitive to modeling errors. The viewpoint adopted is to place a relative entropy bound on modeling errors and formulate estimation problems as minimax problems. In the estimation and filtering case, the optimum solutions take the form of risk sensitive filters.
Prof. Levy has also investigated the application of multirate DSP and statistical signal processing techniques to the digital correction of analog circuit imperfections. In particular, he has examined the blind calibration and correction of timing and bandwidth mismatches for time-interleaved
ADCs, and the calibration of ADC nonlinearities.
He has also studied signal processing problems, such as precoding, equalization, and space-time array processing for MIMO wireless communications.