Automatic Spike Sorting Based on Robust Clustering

The collected spikes from extracellular recordings usually contain noisy data and outliers, which make it difficult to separate them. A method for spike sorting based on robust clustering is proposed to deal with the problem. The clustering method combines the advantage of fuzzy clustering and robust statistical estimators. The number of dusters is obtained by fuzzy cluster validity. In order to reduce the influence of outliers, the validity index is calculated using the weighting intra-cluster distances. The proposed method is suitable to separate neural spikes in the presence of noisy data and outfiers. The experiment on real data shows its performance.

The collected spikes from extracellular recordings usually contain noisy data and outliers,which make it difficult to separate them.A method for spike sorting based on robust clustering is proposed to deal with the problem.The clustering method combines the advantage of fuzzy clustering and robust statistical estimators.The number of clusters is obtained by fuzzy cluster validity.In order to reduce the influence of outliers,the validity index is calculated using the weighting intra-cluster distances.The proposed method is suitable to separate neural spikes in the presence of noisy data and outliers.The experiment on real data shows its performance.