Noise Reduction Techniques Using Higher-Order Spectra

This project will address the development of noise reduction and narrowband/wideband interference cancellation techniques that are needed to support wideband communication systems. Underscoring this effort are advanced signal processing algorithms based on higher-order cumulants or spectra (HOS) techniques, such as the trispectrum. HOS-based techniques have shown great promise in terms of accuracy, robustness and performance improvement over autocorrelation (or power spectrum) -based methods in the face of statistical uncertainties of communication environments. Consequently, the development and application of HOS-based algorithms to noise and interference reduction problems that arise in wideband communications has become timely. Four directions are covered during this exploratory development: (1) utilization of HOS-based Time-Frequency Distributions for the identification and cancellation of narrowband/wideband interferences; (2) development of HOS-based methods for the suppression of additive Gaussian noise (generally colored and possibly spatially correlated); (3) evaluation of the performance of the HOS-based methods against varying SNRs, SIRs, as well as different noise/interference characteristics; (4) development of a Reduced Complexity Algorithmic Architecture (RCA2) and assessment of algorithmic complexity.