Journal of Ambient Intelligence and Smart Environments

Formal modeling and analysis of human behavior can properly advance disciplines ranging from psychology to economics. The ability to perform such modeling has been limited by a lack of ecologically-valid data collected regarding human daily activity. We propose a formal model of indoor routine behavior based on data from automatically-sensed and recognized activities. A mechanistic description of behavior patterns for identical activity is offered to both investigate behavioral norms with 99 smart homes and compare these norms between subgroups. We identify and model the patterns of human behaviors based on inter-arrival times, the time interval between two successive activities, for selected activity classes in the smart home dataset with diverse participants. We also explore the inter-arrival times of sequence of activities in one smart home. To demonstrate the impact such analysis can have on other disciplines, we use this same smart home data to examine the relationship between the formal model and resident health status. Our study reveals that human indoor activities can be described by non-Poisson processes and that the corresponding distribution of activity inter-arrival times follows a Pareto distribution. We further discover that the combination of activities in certain subgroups can be described by multivariate Pareto distributions. These findings will help researchers understand indoor activity routine patterns and develop more sophisticated models of predicting routine behaviors and their timings. Eventually, the findings may also be used to automate diagnoses and design customized behavioral interventions by providing activity-anticipatory services that will benefit both caregivers and patients.

The way people interact in daily life is a challenging phenomenon to be captured and studied without altering the natural rhythm of the interactions. We investigate the development of automated tools that may provide information to the researchers that analyse interactions among humans. One important requirement of these tools is that should not interfere with the subjects under observation, in order to avoid any alteration in the subject’s normal behaviour. Our approach is based on the detection of proximity among groups of people that is obtained using commercial wearable wireless tags based on Bluetooth Low Energy (BLE) and a novel algorithm called Remote Detection of Human Proximity (ReD-HuP) that analyses the wireless signal of tags and produce the proximity information. The algorithm, which has been validated against the ground truth of an experimental dataset, achieves an accuracy of 95.91% and an F-Score of 95.79%.

Vehicular networking has gained considerable interest within the research community and industry. The automotive industry is supporting the notion of pervasive connectivity by agreeing to equip vehicles with devices required for vehicular ad hoc networking. Equipped with these devices, mobile nodes in vehicular ad hoc networks (VANETs) are capable of hosting many types of applications as services for other nodes in the network. This research focuses on addressing the challenges of location-dependence, intermittent network connectivity and irregular network traffic flows in unplanned areas for VANETs to host and operate non-safety-critical VANETs services. We assume unplanned areas as the one that lack communication infrastructure and planning. Such areas observe irregular vehicular traffic on the roads as well as on the networks. This research investigates the shortcomings of location-dependence, intermittent network connectivity and irregular network traffic flows and addresses them by exploiting location-dependent service migration over an integrated network in an efficient and cost-effective manner.

Currently, smartwatches are mainly used as an extension of smartphones. However, equipped with various motion sensors, they are also effective devices for human activity recognition, particularly for those involving hand and arm movements. In this paper, we investigate the smoking recognition problem with motion sensors on smartwatches using supervised learning algorithms. For this purpose, we collected a dataset from 11 participants including ten different activities. The dataset includes different smoking variations in four different postures, such as smoking while standing, as well as similar activities, such as eating, and other activities, such as walking. Instead of approaching the problem as a binary classification problem, such as smoking and other, we are interested in differentiating smoking in different postures. Our aim is to explore the parameter space that may affect the recognition process on a large and complex dataset, considering 4 different window sizes and overlaps, 63 different features extracted from each sensor, 4 different sensors, 2 different sensor combinations, 3 classifiers and 10 different activities. Additionally, we analyze the impact of participants’ height on the recognition performance. The results show that, simple time-domain features and the combination of accelerometer and gyroscope sensors perform the best. When we consider the impact of height on the recognition performance, the results show that it does not have a significant effect when all activities are considered, however, it does have an effect on smoking while standing, particularly for participants with a significant height difference than others.

Predicting the footfall in a new brick-and-mortar shop (and thus, its prosperity), is a problem of strategic importance in business. Few previous attempts have been made to address this problem in the context of big data analytics in smart cities. These works propose the use of social network check-ins as a proxy for business popularity, concentrating however only on singular business types. Adding to the existing literature, we mine a large dataset of high temporal granularity check-in data for two medium-sized cities in Southern and Northern Europe, with the aim to predict the evolution of check-ins of new businesses of any type, from the moment that they appear in a social network. We propose and analyze the performance of three algorithms for the dynamic identification of suitable neighbouring businesses, whose data can be used to predict the evolution of a new business. Our SmartGrid algorithm reaches a performance of being able to accurately predict the evolution of 86% of new businesses. In this paper, extended from our original contribution at IEEE InteEnv’19, we further investigate the influence of neighbourhood venues in prediction accuracy, depending on their exhibited weekly data patterns.