A comparative evaluation of various methods for detection of small spikes in data

The detection of relatively large spikes in data can be easily handled using common outlier detection techniques, but that of the smaller ones becomes a non-trivial task because of their inconspicuous nature. One has to utilize various signal enhancement, processing, and thresholding techniques in order achieve a 100% detection accuracy and a low false alarm rate. Seven different spike detection techniques are studied and compared here. Experimental results are presented in details.