Abstract: Although traditional k-nearest neighbor(K-NN) and Bagging can recognize effectively less human activities using WiFi ambient signal, recognition by either alone of the seven states, namely, empty, walking, sitting, standing, sleeping, falling and running, is not ideal. To improve recognition rates, a new algorithm, fusion algorithm, was designed. It significantly outperforms K-NN and Bagging in terms of recognition ratios in both single-subject and multi-subject experiments.

Key words: WiFi ambient signals, human activities, fusion algorithm, multi-subject