Fingerprint Reconstruction for Smartphone Acoustic Based Target Localization Using Compressive Sensing
Year: 2026
Type: article
Abstract: To mitigate the degradation of fingerprint databases and the associated maintenance overhead induced by environmental variations in acoustic fingerprint–based indoor localization systems, this paper proposes a fingerprint sparse reconstruction framework based on Compressive Sensing (CS). Specifically, we innovatively formulate the reconstruction of the fingerprint database as a sparse signal recovery problem, and introduce a Weighted Sequential Generalized K-means (W-SGK) dictionary learning algorithm. This algorithm effectively extracts sparse bases that characterize the global spatial structure of the sound field, even in non-stationary acoustic environments with low Signal-to-Noise Ratios. Furthermore, by integrating a correlation coefficient-driven adaptive step-size search strategy, the proposed framework achieves precise recovery of the complete fingerprint distribution using a limited number of observational samples. Experimental results demonstrate that, with a sampling rate as low as 40%, the reconstructed fingerprint database maintains localization accuracy comparable to that of a fully constructed database. This study validates the feasibility of replacing full-scale resampling with sparse updates, significantly reducing long-term maintenance costs and update cycles, and providing a highly efficient engineering solution for the sustainable operation of indoor localization systems.
Source: IEEE Internet of Things Journal
Authors: Qiaolin Pu, Yuanyuan Yi, Mu Zhou, Yongkun Shen
Institution: Chongqing University of Posts and Telecommunications
Language: English
Subfield: Electrical and Electronic Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: hybrid
Estimated APC paid: $2,645
OpenAlex ID: W7204459129
