Spike Detection Using Cross-Correlation Based Method

The detection of activities (spikes) in extracellular neural recordings is a crucial step to the classification of neural signals. In a high-channel count microelectrode array the recorded action potentials come from different neurons on different channels, therefore the number of spikes can be huge (more than 10000) after a couple of minutes. The impact on computation resources of feature extraction or classification can be significant when spike detection is working with a high false positive ratio. In this paper a cross-correlation based spike detection algorithm is proposed and compared to other spike detection algorithms. The cross-correlation method uses a predefined average template for the base of the detection. Using this method an average efficiency of 94% with 0.3 false positive rate can be achieved at an estimated signal-to-noise ratio (SNR) of 4 dB. The proposed method is compared to 5 widely used spike detection algorithm namely the Non-Linear Energy Operator (NEO), the Power method, Positive, Negative, and Absolute Amplitude Thresholding (AMPP, AMPN, AMPA) using a neural signal generator with 32 channels, variyng SNR and 12 neurons on the channels. It can be concluded, that the proposed method can achieve a far better false positive ratio, than the others, while maintaining accuracy.