Signal denoising via quadratic semi-infinite programming

This paper describes an approach to signal denoising based on semi-infinite programming. This method offers some advantages over traditional discrete-time filtering, including sharper transition bands between passband and stopband, zero phase response, and uniform signal power in the stopband that can be made arbitrarily low. We also describe a method for notch filtering that uses quadratic programming with linear equality constraints. Both techniques are demonstrated to outperform conventional methods based on LTI filtering.