**4. Experiment Design**

In order to evaluate the contexts completely, we need to design the experiments carefully and correctly. First, we will give a detailed definition of walk and how to distinguish walk from other activities. Then, by simplifying the problem of complex and varying contexts, we could make sure our evaluation is reliable. Last, extensive experiments are conducted, and we will show how to set the parameters of data-preprocessing, feature extraction and model training. We also give an interpretation of false positive in step counting.
