Time series processing at the age of big data: any change of paradigm?

Acquiring data is quite easy, at least until it is exploited, when the results usually show the limits due to acquisition errors. At a time where big data sounds the solution to many problems [2], it is important to signify that it is not because you have a lot of data available that its quality may be less important. Data must be reliable, it must be a precise representation of the reality one wants to observe. In this article we will point out the solutions we used in our last realisations to deal with data evaluation, starting from simple time-series, such as a heartbeat monitoring signal, to more complex ones with several heterogeneous signals, such as hyperspectral images or environmental measurements. All imply careful time acquisition as well as precise value estimations. For each situation that will be presented, we will identify the difficulties and find how to overcome them or, at least, diminish their impact. In any case, having a maximum of information in the signal must be a goal and will be crucial to obtain a good discrimination between signal and noise. The goal here is not to present a new algorithm, an improvement somewhere, or an extraordinary object. The goal is to put together little things that are often unconsidered, and bring difficulties and chaos when pushing a new realisation to its end.