Abstract
© Springer International Publishing Switzerland 2014.Neural action potential, named spike, plays an important role in comprehending the central nervous systems. Neuronal spike detection is a technical challenge due to the effect of strong noise and nonstationarity. There are two main problems for almost all conventional spike detection approaches. First, a filtering approach is often followed for pre-processing the data. Selection of the filter parameters is a time-consuming task. To overcome this problem we suggest utilizing empirical mode decomposition (EMD) and a filter whose parameters are selected automatically. The second problem is that the spike detection method is signal dependent and the performance changes considerably when the data changes. To tackle this problem, a novel approach which utilizes the data distribution is proposed. This method exploits the fuzzy set theory to combine a number of spike detectors to achieve a higher performance. The results demonstrate the superiority of the proposed method.