Key-Device Based Place Recognition Using Similarity Measure between IoT Spaces

Abstract

Location, or place, is a key piece of information when providing context-aware services to users in smart environments. Many researchers in the mobile robotics community or context-aware service system sector have suggested various solutions using vision data processing or additional geographical information such as from Google Maps. However, recognizing a place by processing vision and image data brings a huge burden on Internet of Things (IoT) devices. Geographical information such as a building's exact structural configuration may not be available for IoT devices. As the number of IoTs grows rapidly, limitations of existing approaches can be solved with the IoT environment, since IoT devices in close proximity to one another can detect each other easily. In this paper, we propose a novel approach that utilizes Key- Device values and similarity measure between devices to recognize a place. Our proposed approach measures similarity between IoT space input and representative feature set of each known place a priori, and determines the most probable place of the IoT space. We evaluate with Euclidean distance, cosine similarity, and weighted cosine similarity. Experimental results show that the Key-Device concept remarkably improves precision and recall of place recognition.