Skip to main content
. Author manuscript; available in PMC: 2017 Mar 1.
Published in final edited form as: J Neurosci Methods. 2015 Dec 19;261:135–154. doi: 10.1016/j.jneumeth.2015.12.004

Figure 8.

Figure 8

Comparison of time-windowed and frequency-windowed adaptive filters applied to line noise in electrophysiological data. Filtering was applied to 60 Hz line-noise contaminated ECoG data, whose frequency and amplitude varied slightly over time. A: A denoising routine implemented in the Chronux toolbox (rmlinesmovingwinc.m) removed sinusoidal signal components within finite overlapping 12 s segments with 6 s overlap. Significant distortion of neighboring frequencies results from spectral leakage associated with finite windowing. Plots show log power difference for unfiltered and filtered data. B: DBT denoising (bandwidth = 1/6 Hz) avoids spectral leakage artifacts. C: A comparison of power in the noise signals isolated by respective filters reveals much greater energy outside the range of line-noise in the time-windowed method as compared with the DBT approach, a consequence of spectral leakage in the former.