I am looking for a data set of multivariate time series data which contains classes that only require a subset of time series for their identification. Assume a human activity recognition data set with sensors mounted on different body parts. Not for every activity, I need to consider all sensors. For instance, a sensor mounted on the leg is not relevant for detection of a "hand shake".
Does anybody know of such a data set? I need it for research purposes but cant find one. I think in real life data sets there are often difference relevances of attributes for each class. For instance, monitoring a whole system where some classes are only related to smaller parts of the system and other classes to the whole system. I think such data sets are very interesting. For any ideas or data sets, I am very thankful!
Thanks in advance!