An Attention Auxiliary Network-Based Method for WiFi Fingerprint Indoor Localization

The fingerprint location system based on WiFi RSSI is often affected by APs changes, which makes it difficult to obtain positioning signals with high location distinction and long-term stability. However, existing methods for reducing cross-domain fingerprint distribution discrepancy don't effectively introduce specific features of different time domains and cannot achieve long-term accurate localization. In this paper, we proposes a WiFi fingerprint indoor positioning algorithm based on attention auxiliary network(AAN). Specifically, we first extract the specific feature in different time domains and transfer heterogeneous fingerprint to homogeneous fingerprint. Then, we increase the weight of important features based on the attention mechanism. Finally, we extract hierarchical features to improve the location distinction of fingerprints. Experimental results show that AAN achieves state-of-the-art localization accuracy and is robust to long-term changing environments.