Interference Mitigation in Spectrum Sharing Environments Using Time- Frequency Processing and Feature Clustering

This paper considers a radar tracking problem in an environment where spectrum is shared with a multiuser wireless communication system. At the radar receiver, the large number of communication transmissions interfere with the noisy and time-delayed transmit signal. We propose to improve the radar estimation performance by integrating a highly-localized time-frequency method with unsupervised clustering. The method first uses the iterative matching pursuit decomposition (MPD) algorithm to represent the overall received signal as a weighted linear combination of highly localized Gaussian waveforms. The use of the MPD can help reduce the high noise in the received signal by appropriately selecting the algorithm's stopping criterion. The MPD provides low-dimensional features that can be used to cluster the multiple signal components using Gaussian mixture modeling. The proposed estimation method integrates along lines of a modified Wigner distribution of the clustered components in the time-frequency plane. Using simulations, the radar estimation performance is shown to improve when combining the MPD with Gaussian mixture modeling.