A nonlinear lowpass filter that eliminates peak attenuation

Linear-phase lowpass filters are often used as a preprocessing step in signal processing applications to eliminate high frequency components of noise that do not overlap with the signal spectrum. Signals with sharp peaks confound this type of noise reduction because sharp peaks have a broad spectrum and are significantly attenuated by lowpass filters. This paper describes a simple method that restores the signal peaks to their full amplitude by adding a filtered estimate of the peak residuals to the original lowpass filter output. Examples are given of the nonlinear filter applied to a Poisson process, an electrocardiogram, and an extracellular microelectrode recording of neuron action potentials.