**3. WD and SC Algorithms**

Walk detection (WD) and step counting (SC) algorithms were generally designed following a similar routine: (1) feature extraction; (2) feature detection; and (3) state recognition. In an implementation, the design methodologies can be roughly categorized into: (1) heuristic-based; (2) signal processing based; and (3) machine learning-based. The related algorithms are briefly summarized and introduced.

Since we aim to give a comprehensive evaluation of various contexts, a proper set of algorithms that could exactly reflect the contexts' changes should be selected. Some features and algorithms that are too sensitive or not widely applied could not provide fair comparisons of different contexts. Too complicated features and some 'best' algorithms should be avoided because they may generate biased results. Therefore, we surveyed many papers on both feature extractions and classification algorithms and selected the most stable features and algorithms to conduct the experiments and evaluations.
