Detection of Neural Spike in Extracellular Recording Using Shannon Energy and Hilbert Transform

Neural spike detection is the first step in the analysis of neural action potentials in extracellular recordings. The background noise which mainly originates from a large number of far neuronal units, usually confront with detection of low-amplitude spikes. So far, many scholars have devoted their works to this subject and many algorithms have been proposed. In this paper we present an automatic spike detection algorithm for the noise-contaminated extracellular signal. This algorithm consists of four steps: 1- A bandpass filtering and using a differential filter; 2- applying Shannon's energy nonlinear filter; 3. Hilbert transform; and 4- Thresholding of the signal. The proposed method has been compared with five known methods in spike detection. This comparison is done on two simulated datasets and one real data set. The results indicate the superiority of the proposed method for simulated data compared to other methods, which indicates the robustness of the proposed algorithm to the noise. Meanwhile, for real data, it reaches the second place among all six methods. Using Shannon's non-linear energy filter can be an effective way to detect spikes in extracellular signal recordings. The comparison indicates that this method is superior to the commonly known methods for spike detection.