POSP

In the context of bio-signal modeling, unsupervised techniques that highlight recurring patterns permit to assess the local structure of the signal but fail to capture the global one. With POSP, I propose to develop efficient tool that couple such techniques with Point Processes to account for the global temporal dependencies of non-stationary signals, in particular for non-stationary pseudo-periodic signals. The developed methods will be applied to neurophysiological signals in order to reveal temporal brain areas connections and transient oscillatory phenomenon.