Class Survival
When dealing with real data, it is expected that some blocks could be dominated by noise or artifacts, so the contribution of neurons could be masked. To handle this issue, a survival criterion was added to the method. Under this condition, any class would survive for a maximum of \(CS=3\) consecutive blocks, even if the tracking method is not able to associate clusters to that class in the intervening blocks. Within these blocks, the same overlapping set for the class is used (as no new spikes are incorporated). If no spikes are assigned to the class after \(CS\) consecutive blocks, the class is labeled as "disappeared", and it is no longer tracked.