Optimal Spike Detection Technique Based on Amplitude Threshold

Abstract

Listen

Translate article

Spike detection as the first basic step is very important in analysing and classification of data, since the quality of resulting data depends crucially on a particular detection technique used. This paper presents the optimal spike detection technique based on amplitude threshold on a band pass filtered signal. Comparisons were made on four different filtered signals: voltage of the entire signal, power of the entire signal, voltage moving average and power moving average of the signal. MATLAB was used to generate six different realistic simulations with varying signal to noise ratio which resembles that of a real dataset. The six different simulations contain ten samples each. For each simulated signal, only one type of spike shape was used with same firing frequency following a Poisson distribution. The duration of the simulation was equal in all cases with the signal to noise ratio defined as the amplitude of the spikes normalized by the noise level. Also, the threshold for spike detection was calculated based on the estimation of the standard deviation of noise; the area under the receiver operating characteristic curve and statistical analysis of data were used to quantify their performance. The major finding is that the voltage technique superseded all other techniques mentioned above both in high and low signal to noise ratio. When voltage was compared with power, voltage moving average (vma) and power moving average (pma), it was observed that p (0.0022) 0.05 with h = 0, also implying that the test accepts the null hypothesis of equal medians. For comparison of vma and pma, it was observed that p (0.6991) > 0.05 with h = 0 which presents that the test accepts the null hypothesis of equal medians. However, is clear that voltage technique is the optimal spike detection technique based on amplitude threshold.