Spike Sorting by Joint Probabilistic Modeling of Neural Spike Trains and Waveforms

Table 4

Classification error rates (FP + FN) for the HC1 dataset.

Dataset	SNR	GMM	WaveClus Method		Proposed
			Wavelets	PCA
Dataset 1	Clean	10.64%	32.02%	17.98%	5.60%
	15 dB	9.73%	24.5%	16.79%	4.77%
	10 dB	9.73%	31.84%	21.47%	4.95%
	5 dB	11.28%	31.93%	32.02%	6.51%
	0 dB	10.19%	32.11%	31.56%	9.27%
	−5 dB	20.09%	32.11%	32.02%	15.23%
	−10 dB	31.93%	67.98%	31.93%	32.11%
Dataset 2	Clean	2.09%	8.49%	6.56%	1.86%
	15 dB	1.96%	7.56%	6.20%	1.76%
	10 dB	1.99%	7.59%	6.27%	1.89%
	5 dB	2.29%	19.62%	11.04%	2.06%
	0 dB	3.51%	19.42%	14.29%	3.45%
	−5 dB	6.99%	19.65%	19.52%	5.87%
	−10 dB	32.55%	19.59%	19.56%	30.53%