DESPIKING DISCRETE-TIME SIGNAL USING HISTOGRAM METHOD
Time series may contain undesired transients and spikes. This function replace spikes (outliers) exceeding the threshold value by interpolating among previous and subsequent data points or replacement them with NaN per user choice. The threshold is defined as mean +/- a number of standard deviations of windowed data centered at spike locations. This code uses the histogram method of Solomon et al. (2001) to detect spikes.
Examples are given in the comment section. The debug option when it is "True" compares results of this function with MatLAB's 1-D median filtering (medianfilt1) function.
For questions, comments and suggestions: email@example.com
Solomon, O. M., D. R. Larson, and N. G. Paulter (2001). Comparison of some algorithms to estimate the low and high state level of pulses, IEEE Instrumentation and Measurement Technology Conference, Vol. 1, Budapest, Hungary, 21–23 May 2001, 96–101.
Dr. Erol Kalkan, P.E. (2019). spikeRemoval (https://www.mathworks.com/matlabcentral/fileexchange/69614-spikeremoval), MATLAB Central File Exchange. Retrieved .
Replaced diff with nth order central difference function.
Added return -1 if no spike found
Added output array for indices of spikes, their values and replacements (thanks to Dr. Ayal Anis)
Comments are updated
Excluded spike in spike search window for mean and standard deviation computation; bug fix for spike search in the beginning and end regions of the input array.
Reference added, description updated
Code re-organized to eliminate the wrapper.