LTPDA Toolbox™

Spikes reduction in data

Description

Spikes in data due to different nature can be removed, if desired, from the original data. LTPDA spikecleaning detects and replaces spikes of the input AOs. A spike in data is defined as a single sample exceeding a certain value (usually, the floor noise of the data) defined by the user:

where is the input data high-pass filtered, is a value defined by the user (by default is 3.3) and is the standard deviation of . In consequence, a spike is defined as the value that exceeds the floor noise of the data by a factor , the higher of this parameter the more difficult to "detect" a spike.

Syntax

    

    b = spikecleaning(a, pl)

Parameters

The following parameters can be set in this method:

'kspike' - set the value (default is 3.3)
'method' - method used to replace the "spiky" sample. Three methods are available ---see below for details---:

'random'
'mean'
'previous'

'fc' - frequency cut-off of the high-pass IIR filter (default is 0.025)
'order' - specifies the order of the IIR filter (default is 2)
'ripple' - specifies pass/stop-band ripple for bandpass and bandreject filters (default is 0.5)

Algorithm

Random: this method substitutes the spiky sample by:

where is a random number of mean zero and standard deviation 1.

Mean: this method uses the following equation to replace the spike detected in data.

Previous: the spike is substitued by the previous sample, i.e.:

Examples

1. Spike cleaning of a sequence of random data with kspike = 2.

      

      x = ao(plist( 'waveform', 'noise', 'nsecs',1e4, 'fs',10)); % create an AO of random data sampled at 1 Hz.
      pl = plist( 'kspike', 2);                         % kspike = 2
      y = spikecleaning(x, pl);                         % spike cleaning function applied to the input AO, x
      iplot(x, y)                                       % plot original and "cleaned" data

Interpolating data

Data gap filling