InferLoc: Calibration Free Based Location Inference for Temporal and Spatial Fine-Granularity Magnitude

Abstract

Location is the most important information in the field of context-aware computing. Normally, one location represented as absolute physical coordinate is less understandable than semantically meaningful place like "home", "office", etc. This paper proposes a novel calibration free based algorithm called Infer Loc to infer user's daily significant locations using Wi-Fi signals obtained from mobile phone. Infer Loc contains three main steps: 1) Stop point detection based on trajectory segmentation through similarity calculation between neighbor sampling windows, 2) Location discovery through density based clustering and 3) Semantically significant location inference through matching clustered locations and recorded places in personal diary. Furthermore, we implement and validate Infer Loc algorithm on realistic data collected from real-world wireless environment. Experimental results show that Infer Loc can recognize visiting locations both in temporal and spatial fine-granularity magnitude under short response delay.