Conclusion

In this article, we present a state-of-the-art procedure for recognizing activities from accelerometric data. This procedure is structured in different stages, from data collection to the design of a classifier and the validation or otherwise of the entire procedure. The implementation of these different steps induces a number of choices by the user and, although there is no consensus on many of these choices, this article proposes a selection of the usual choices from the literature.

In the light of the various projects presented in the literature, it would appear that the successful implementation of the activity recognition procedure is conditional, on the one hand, on a good understanding of the problem posed (what activities are to be recognized? what will be the use of this activity recognition?) and, on the other hand, on a good command of the technological...