Extraction and selection of classification parameters
Activity Recognition from Accelerometer Data

Accelerometer data can be used directly by the classification algorithm. For example, data from all three axes of the accelerometer are used directly in . However, in most cases, a set of parameters is calculated from the window data. This set of parameters forms a vector which is then used for classification. The aim here is to select the parameters that best discriminate between different activities.

It is possible to use a large number of parameters or parameter combinations, but it is preferable to use as few as possible, as using inappropriate or redundant variables...