Identification of Social Aspects by Means of Inertial Sensor Data

Bedogni, Luca; Cabri, Giacomo · 2020 · Crossref

DOI: 10.3390/info11110534

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This paper addresses the challenge of identifying user travel paths and social habits without relying on GPS, which requires explicit user permissions and suffers from accuracy limitations. The authors propose CERT (Computationally Efficient algorithm to Reconstruct Vehicular Traces), a novel method that reconstructs vehicular trajectories using only inertial sensor data—specifically accelerometers and magnetometers—available in smartphones and vehicles. The motivation stems from the desire to provide context-aware services while mitigating the computational inefficiencies and error accumulation associated with traditional Dead Reckoning Systems (DRS). CERT aims to match sensor-derived paths to real-world road networks using publicly available OpenStreetMap data, offering a privacy-preserving alternative to GPS-based tracking. The methodology employs a graph-theoretic approach involving two graphs: $G(P)$, constructed from raw sensor measurements, and $G(I)$, derived from OpenStreetMap data. $G(P)$ is built by identifying road segments through magnetometer variance analysis and calculating segment lengths and angles via a DRS using accelerometer data. To ensure computational efficiency, CERT pre-computes N-clique subsets from $G(I)$ for a given area, allowing for rapid matching against sensor data rather than traversing the entire graph for each detection. The algorithm matches $G(P)$ to subgraphs in $G(I)$ by minimizing the difference in edge lengths and angles, weighted by parameters $\alpha$ and $\beta$ determined by the statistical variance of road geometries in the specific urban landscape. The study validates the algorithm using a custom Python simulator that generates synthetic inertial data for random paths across various global cities, incorporating realistic sensor noise levels. The results demonstrate that CERT achieves near-perfect path identification in small and medium-sized cities, where road layouts are more unique. The study finds that detection accuracy improves with path length, noting that paths with at least 10 turns can be uniquely identified in most cities worldwide, including large metropolitan areas. However, urban dynamics significantly influence performance; cities in Europe and Oceania are more easily identifiable due to varied road angles, whereas cities in the Americas, characterized by grid-like "Manhattan" layouts with similar turns and road lengths, exhibit greater resilience to this tracking method. The algorithm’s computational complexity is optimized by performing the NP-hard clique generation offline, resulting in linear-time matching during detection. The significance of this work lies in demonstrating that inertial sensors alone can reliably reconstruct vehicular traces, raising important privacy concerns regarding the potential for tracking users without their knowledge or consent. The findings highlight that urban design plays a critical role in the feasibility of such surveillance, with grid-based cities offering more privacy protection than irregularly laid-out ones. CERT provides a computationally efficient framework for trace reconstruction, contributing to the fields of context-aware computing and privacy research by quantifying the risks associated with ubiquitous inertial sensors.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• telematics crash prediction