A Fusion Method for Estimating the Walking Direction of Smartwatch Users

Accurately estimating the walking direction of smartwatch users is critical for applications such as exercise trajectory analysis.This study introduces a novel approach that fuses estimation direction from inertial sensors and GPS.Inertial sensors provide stable estimates as they are unaffected by environmental conditions, but their accuracy can be impacted by sensor performance and user motions, such as hand swinging.GPS, in contrast, offers higher accuracy than inertial sensors under favorable signal conditions.To leverage the strengths of both sensors, the proposed method employs an information-weighted consensus filter, integrating direction estimates and error covariances.Experimental results demonstrate that the fusion approach reduces estimation errors compared to individual sensors.