COPELABS DSpace Repository

The Role of Smart Data in Inference of Human Behavior and Interaction

Show simple item record

dc.contributor.author Sofia, Rute C.
dc.contributor.author Carvalho, Liliana I.
dc.contributor.author Pereira Melo, Francisco
dc.date.accessioned 2018-04-06T10:17:46Z
dc.date.available 2018-04-06T10:17:46Z
dc.date.issued 2018-04
dc.identifier.citation Rute C. Sofia, Liliana I. Carvalho., Francisco de Melo Pereira, Samrat Dattagupta.The Role of Smart Data in Inference of Human Behavior and Interaction. Book chapter. "Smart Data: State-of-the-Art and Perspectives in Computing and Applications". Editors:K.-C. Li, Q. Zhang. L. T. Yang, B. Di Martino. CRC Press, Taylor & Francis en_US
dc.identifier.isbn 1138545589
dc.identifier.uri http://hdl.handle.net/20.500.11933/746
dc.description.abstract Pervasive computing has been mostly used to build systems encompassing a small number of devices that interact with single users or small groups. As technology becomes truly pervasive, the growth of low-cost sensing systems that are built based on regular personal devices has exploded. Such devices are carried around by a large number of people, contributing to large-scale highly variable sensing environments. Today, such large-scale scenarios are often denoted to be associated with urban scenarios, where activity recognition often derived from a centralized architecture is applied to forecast the most varied aspects, ranging from traffic jams to whether or not a restaurant may have the attendance conditions expected. Several attempts to model different aspects of human activities, social interaction, and roaming patterns have been increasing, having as common goal to assist large-scale sensing in reaching a level where activity recognition can integrate, in a non-intrusive way, cognitive aspects also. Such attempts carry several challenges from a software and networking architecture perspective for instance: light methods for accurate sensing; capability to extract data and to infer future activities based on small amounts of extracted data; not infringing the privacy of the involved citizens, by keeping captured data local and hence, recurring whenever feasible to distributed inference [#Lane11:coco, #Reiche08:music]. A second relevant challenge to overcome is to ensure that the technology is truly non-intrusive, i.e., does not rely on the explicit input of information from citizens. This means that data needs to be captured in a seamless way that does not jeopardize the citizen's privacy or anonimity. And that data must be kept local so that only the user has access to it. A third relevant challenge relates with the capability to take advantage of opportunity to perform sensing. Opportunistic sensing[#Lane08:urban, #Sofia2016] ensures that the owner of sensorial devices remain agnostic of any sensorial activity: the device is activated whenever its state matches the requirements of a sensing application, and the latter does not have an impact on user experience. However, opportunistic sensing occurs without user intervention and may be required to infer about complex activities. Hence, a fourth and major challenge to address is the correlation of contextual conditions and personal characteristics (e.g., age and lifestyle) encountered in large-scale mobile sensing systems, as well as the integration of multiple data representing the same real-world activity into a consistent, accurate, and useful representation [#MeloPereira2016]. This chapter explores features, concepts, and provides guidelines concerning the role and applicability of smart data captured in a non-intrusive way, in the inference and contextualization of human behavior and interaction [#Sofia2015][#Carvalho2017, #Kofod-Petersen2006]. The chapter starts by introducing aspects related to human interaction, for instance, how to define and to best model physical and psychological proximity; models for social awareness and social contextualization. The next part of the chapter deals with interaction inference and interaction contextualization, namely: classification models that best suit the inference of behavior in the vere of smart data; challenges in regards to small data capture and behavior inference derived from small data, in particular when considering decentralized, mobile cyber-physical systems; guidelines to model interaction based on pervasive wireless sensing systems, including available middleware and systems (tools). The chapter then provides information concerning specific applicability use-cases, namely, Points of Interest detection via smart data, and how smart data can be used to boost social interaction in a pervasive, non-intrusive way. The chapter concludes with a set of recommendations. en_US
dc.description.sponsorship This work has been developed under the Fundação para a Ciência e Tecnologia PDLAB project UID/MULTI/04111/2016. en_US
dc.publisher CRC Press, Taylor and Francis, USA en_US
dc.subject smart data, behavior inference en_US
dc.title The Role of Smart Data in Inference of Human Behavior and Interaction en_US
dc.type Book chapter en_US


Files in this item

Files Size Format View

There are no files associated with this item.

This item appears in the following Collection(s)

Show simple item record

Search DSpace


Advanced Search

Browse

My Account