Skip to main content
. 2017 Mar 30;12(3):e0174790. doi: 10.1371/journal.pone.0174790

Fig 3. Filtration effect on spike shape in simulated data.

Fig 3

(A) The effect of the spike width, defined at the half height of the waveform, on the distortion: Top—Examples of waveform distortions for different simulated waveform widths. Bottom—distance between the filtered and raw waveforms as a function of the waveform width. (B) Spectral composition of the waveforms referred in (A). (C-D) The effect of the cutoff frequency on the distortion of the waveform. (C) Mean action potential waveforms taken from an identical signal filtered with different filter cutoff frequencies (blue color scale—NLP, red color scale—ZP, black—raw waveform) (D) Distance between the filtered and raw waveforms as a function of the filter's cutoff frequency. (E-F) Effect of filtration on the detection of action potentials: (E) ROC curve of the detection of spikes using different thresholds for NLP and ZP (SNR = 2.7) filtered signals. (F) Detection error—the area under the ROC curves as a function of SNR. (red—ZP, blue—NLP).