Automatic detection of scan periods

[tsalo]

Detailed Description

In cases where triggers are not used or are accidentally dropped, it could be useful to try predicting when scans are happening based on patterns in the signal. Some signals, like EDA and ECG, are dramatically affected by the pulse sequence, so we should be able to infer onsets and offsets with fairly high accuracy if those modalities are recorded.

Context / Motivation

The goal is to make it easier to convert data with missing trigger information.

Possible Implementation

I imagine that the implementation would be similar to the classification in #204, except applied to predicting time points/chunks in time series rather than classifying whole time series.