Classification of neural spike under nearly 0 dB signal-to-noise ratio

We present neural spike sorting when the signal-to-noise ratio (SNR) is close to 0 dB. The use of nonlinear energy operator enables detection of an action potential even when the SNR is so poor that the usual amplitude thresholding method cannot be applied. Thus training sets that effectively represent the probability distribution of the input vectors can be obtained and the learning capability of the neural network classifiers can be better utilized The trained classifiers exhibit correct classification ratio higher than 90% when the SNR is as low as 1.2 (0.8 dB) when applied to the extracellular recording obtained from Aplysia abdominal ganglion using semiconductor microelectrode array