A Probabilistic Place Extraction Algorithm Based on a Superstate Model

Research on place extraction has been of interest for the detection of meaningful places that users visit. Because interpretations of meaningful places may be different according to location-based applications, a universal place extraction algorithm that is able to detect all kinds of meaningful places needs to be developed. Unfortunately, most previously proposed place extraction algorithms failed to show high place detection accuracy and also failed to perfectly detect meaningful places. In this work, we propose a new place extraction algorithm that can significantly enhance the accuracy of place extraction. The basic concept of the proposed algorithm is a superstate model, which is an extension of the Hidden Markov Model (HMM); we substituted superstates for the simple probabilistic distributions of the HMM. Our proposed algorithm shows remarkable detection accuracy in place extraction, significantly higher than any other previously proposed algorithms. Furthermore, the proposed algorithm can efficiently operate in mobile environments because its computations are simple.