Knowledge-assisted location-adaptive technique for indoor-outdoor detection in e-learning

Abstract

The era of pervasive and ubiquitous computing has brought the learning far beyond the traditional classrooms to distant and mobile e-learning. Being easily accessible through time and place, e-learning systems rushed into masses and quickly appeared under the criticism as being uni-directional and fitting various learners under “one size”. In order to differentiate learners' needs and to apply the most suitable educational approach to the particular learner, researchers have introduced the Learner's context - a set of preferences defined by the learner's personal characteristics, technical capabilities of the user device, and the environment where learning takes place. Concerning the physical learning environment, the basic requirement is to distinguish between indoors and outdoors (IO). Existing approaches for IO-detection either apply pre-defined hard-coded thresholds to the sensing parameters or use machine-learning techniques. While the latter demonstrates a more adaptive approach for IO-detection over the former, decisions based on training data are not accurate once the environment is significantly changed, which is highly relevant for the modern learner with increased mobility. In this paper, we propose a novel knowledge-assisted location-adaptive technique for IO-detection in e-learning scenarios. The technique leverages data collected from various ambient sensors such as light, temperature, humidity, and noise and compares them with characteristics that the e-learning environment has at this point in time and in the current physical location being inside or outside. Here, we model the e-learning environment based on the empirical observations of the natural learning process augmented by the knowledge about current weather and environmental conditions collected from the weather web-service. The proposed approach is easily adaptable to the changing conditions in time and place with no need for the training phase. This work can be the first step towards robust location-adaptable IO-detection algorithms.