Tracking Loop
The tracking loop is applied after a sorting algorithm classified the spikes of a given block \(t_i\), which includes all the sets \(OS\) associated to the classes from the previous block \(t_{i-1}\). Following the calculation of the overlapping value \(Ov_{j,k}^{t_i}\) for every pair \(j,k\), each cluster \(\tilde C_k^{t_i}\) will be associated to one of the following cases:
- The cluster is labeled as a "match" if it has an overlapping value higher than 0.5 against a single class \(j^*\) from the previous block. In this case, all the spikes from the cluster are incorporated into the class \(j^*\) (see Fig. \ref{595195}A).
- The cluster is labeled as a "split" if its highest overlapping value against all classes is between 0.2 and 0.5, and the second highest value is less than 50% of the highest and lower than 0.2 (see cluster \(\tilde C_1^{t_2}\) in Fig. \ref{595195}B).
- The cluster is labeled as a "merger" if it has overlapping values higher than 0.5 with more than one class. In this case, the procedure explained below in \ref{108623} is applied (see Fig. \ref{595195}D).
- In any other case, the cluster is labeled as a "new class" (see cluster \(\tilde C^{t_2}_3\) in Fig. \ref{595195}C).
If the sum of the overlapping values from a given class \(j'\) for all the clusters labeled as split or match is higher than 0.5, the spikes from the split clusters are incorporated to the class \(j'\) (see clusters \(\tilde C_1^{t_2}\) and \(\tilde C_3^{t_2}\) in Fig. \ref{595195}B); otherwise they are assigned to a "new class" (see cluster \(\tilde C^{t_2}_1\) in Fig. \ref{595195}C). Using this criterion it is possible to fix cases where the sorting algorithm generates overclustering on a given block.