FT-Loc: A Fine-Grained Temporal Features-Based Fusion Network for Indoor Localization

• DOI: 10.1109/jiot.2023.3299625
• Publication Date: 2023-07-28T17:35:26Z
• Authors: Minmin Liu; Xuewen Liao; Zhenzhen Gao; Qiao Li
• Keywords: Discriminative model; Fuse (electrical); Representation; SIGNAL (programming language); Feature Learning

Indoor location-based services (LBS) are critical for enhancing social and commercial activities that require accurate efficient localization techniques. Existing deep-learning-based indoor methods mainly focus on pre-defined global features to learn local discriminative representations, which increases learning difficulty is not or robust scenarios with small variations. To address the above issues, we propose a novel Fine-grained Temporal based Localization (FT-Loc) framework utilizes multiple sub-signal provide location estimation, each represents piece of clue specific position. Specifically, proposed takes signal sequences as input, Deep Networks considering temporal correlations designed extracting from corresponding clues, respectively. Then, lightweight attention generation scheme used importance representation. Guided by obtained values, fuse generate more distinguishing ones localization. The experimental results show FT-Loc significantly outperforms existing schemes accuracy improvements at least 43.36%.