Enhancing User Experience with Brain-Computer- Interfaces in Smart Home Environments Pierluigi Casale Member IEEE, Juan Manuel Fernández, Xavier Rafael Palou, Sergi Torrellas, Mojdeh Ratsgoo and Felip Miralles Barcelona Digital Technology Centre, C/ Roc Boronat, 11, MediaTIC Building 08018 Barcelona, Spain {plcasale, jmfernandez}@bdigital.org Abstract - In this work, the benefits of Ambient Intelligence for enhancing user experience with Brain Computer Interfaces are explored. In a smart-home environment, statistics of devices activations are used to learn user habits and to adapt the interface for providing the most usual option to the user, reducing the time spent navigating through hierarchical menus. The activation statistics are learned by discriminative machine learning algorithms able to provide the most suitable options for the user interface. Promising experimental results on simulated scenarios encourage following on this research direction. Keywords: Ambient Intelligence, Smart Environments, Brain Computer Interfaces, Statistical Machine Learning. I. INTRODUCTION Ambient Intelligence (AmI) [1] is a new computational paradigm aiming to provide a significant beneficial influence to our society. AmI states on the idea that surroundings are enriched with fully interconnected technological devices and an intelligent system provides proper information to support people in their daily lives with relevant services personalized on their needs. There are many areas of Computer Science such as Embedded Systems and Ubiquitous Computing which contribute to AmI. Nevertheless the predominant role of Human-Computer Interaction and Artificial Intelligence makes possible the AmI envisioned goal, guaranteeing the minimum human involvement into the technological complexity while enhancing user interactive experience. Under this framework, smart homes and intelligent environments found its direct field of application for AmI, providing a new framework for safe environment where people with special needs can improve their physical autonomy and social inclusion. One example of how the promises of AmI can become soon a reality is the BrainAble project [2]. This project is intended to empower people with severe motor disabilities and pursues to mitigate the limitations they suffer in their everyday life. With this goal in mind, the aim of BrainAble is to research and design a Brain Computer Interface (BCI)-based system that improves the interaction between users and their surroundings. Nevertheless, BCIs need that the user must constantly pay attention for interacting with the interface. This issue is an obstacle for the use of these systems in everyday life. Users will be engaged in a continuous control state, their distractions cause misclassifications and the speed of selection will not take into account users’ current psychophysical condition. This paper presents the first steps towards enhancing user experience in BCI-based system by means of AmI. The goal is to reduce the number of interactions needed to perform the user desired action using an intelligent system based in the AmI paradigm. The system is able to learn the statistics of activation of the elements in the environment and to present to the user the most suitable option depending on the context and the learned habits. Using this methodology, the system is able to adapt the BCI interface with the most suitable option for the user and to offer shortcuts to the most used options, reducing the time that the user spends navigating through hierarchical menus. These suggestions are based on predictions made by a set of discriminative machine learning classifiers which learn the statistics of home services usage, the environment and the interaction with social tools. The prediction provided by the classifiers is translated into the presence of the option of activating such element in the BCI interface. Using simulated scenario and statistical validation techniques, it is possible to provide quantitative measure about the performance of the system already in the first stage of development. This paper is organized as follows. Section II presents the usability issue in BCI interfaces. In Section III, the role of AmI for helping in the solution of the considered problem is described. Section IV presents the validation methodology used and, in Section V, results obtained on simulated data are presented with related discussion. Finally, Section VI concludes the paper. II. USABILITY ISSUES IN BCI USER INTERFACES BCIs enable human-computer interfaces controlled with the only cerebral activity, i.e. without muscular intervention. This form of interaction allows persons with heavy neuromuscular handicap to benefit the use of ICT systems. Invasive and non- invasive techniques can be distinguished for sensor cerebral activities, the former using signal coming from surface electroencephalographic record, the latter using the signal coming from deeply implanted electrodes on the brain. Although invasive techniques allow better quality enabling complex applications, they require heavy surgical intervention planning strong ethical issues. The non invasive method of P300 speller is based on event related potential which are 2012 Eighth International Conference on Intelligent Environments 978-0-7695-4741-1/12 $26.00 © 2012 IEEE DOI 10.1109/IE.2012.71 307 2012 Eighth International Conference on Intelligent Environments 978-0-7695-4741-1/12 $26.00 © 2012 IEEE DOI 10.1109/IE.2012.71 307 2012 Eighth International Conference on Intelligent Environments 978-0-7695-4741-1/12 $26.00 © 2012 IEEE DOI 10.1109/IE.2012.71 307