A I R Lab
Ambient Intelligence Research lab
Application Domains and Projects:
Novel application domains result from the integration of technology and user knowledge:
- Aging-in-place: Well-being, health and comfort
While accident detection for the elderly has been identified as a canonical well-being application, serving the user in a smart environment goes beyond implementing a fall detection algorithm, and can include monitoring all the activities of the user in a user-controlled setting. An Ambient Intelligence system can accumulate knowledge from observing the user and provide helpful instructions to the user to promote healthy activities. Several research challenges can be identified in creating such applications. Integration of sensed data from heterogeneous sources as well as with the saved history of the user offers challenges on the methods for combining, validating, cross checking, interpreting, and evaluating the fusion results. The evaluation can be performed through a query from the user, confirming with an alternative expert system, or via demographic comparisons with the relevant data on the internet. Much needed research can hence be defined in handling input from heterogeneous information at the sensor or semantic levels in this application domain.
- Smart buildings: Occupancy awareness and energy efficiency
Observations of the behavior of users in their living or working environments with respect to energy usage and abstracting models can lead to delivering an accepted level of service to the users while reducing energy costs when possible. Related research issues are acquiring contextual information from the user both in real-time in terms of the user's location and activity, and from the history and user profiles, and methods to make decisions based on such contextual data. New sensing and interfacing technologies as well as novel appliances and devices such as wirelessly controlled lights, handheld user devices acting as home control interfaces, and heterogeneous media delivery tools can serve to drive applied and hands-on research in this application domain.
- Meetings of the future
Meeting and seminar rooms are equipped with multiple cameras which either provide close-up views of the participants or views of the interesting spots such as speaker's podium and whiteboard. In teleconference meetings, a remote participant may be interested in receiving a select view of the meeting based on the events that occur. The process of best view selection can be made dynamically based on the contents of all local views. For example, when a participant starts to speak, the view from the tabletop camera which shows that person could be selected for streaming to the remote participant. In addition, different remote participants may have different preferences about the definition of a best view. As an example, a remote participant may prefer to see the view of a seminar's speaker while another viewer may prefer to monitor the expressions of the audience during the seminar. An adaptive system can learn the preferences of each viewer and offer them best views dynamically accordingly.
Analyzing the content of the views that monitor the events of a meeting can also lead to the production of a meeting report. Such reports can be used by the participants to analyze their own performance in the meeting, or by a meeting manager to study the participation of the attendees and to evaluate whether the meeting was balanced in terms of the discussion flow. Furthermore, a report can include a list of interesting events in the meeting with their timestamps. Such a list can be used to annotate the stored video feeds from the meeting so offline viewers can run searches and select the parts interesting to them.
- Smart phone sensor fusion
Smart phones have become a pervasive interface device connecting their users to the world of information. The interface role is indeed a three-pronged system, one being the connection to the digital world on the internet, the second interface being the HCI role which constantly registers the interactions of the user and potentially the user's choices and preferences in various applications, and the third interface being the connectivity that the smart phone offers with the physical world around the user measured by various sensors the user carries or installs in their home or office spaces.
There is a significant opportunity to develop a variety of applications based on data collection and interactions between a user's smart phone, its internal sensors, and other sensory devices around the user, and the vast amount of information available on the internet. Methods of data fusion, preference learning, and behavior modeling are applicable to develop personalized and adaptive services for the user, who in this case carries the smart phone almost at all times.
- Pervasive social networks
One of the representation choices for users in a social network setting is avatar-based communication with others, where a user's physical or emotional state is conveyed to other users through the appearance of an avatar. For example, a user's physical activities or location inside the home can be observed by a camera network, analyzed and mapped to a virtual space in a social network setting in which an avatar represents the user by repeating the user's actions. The emotional state of the user may also be monitored and mapped to the avatar. The user may prefer not to share some of the attributes or change the way the avatar appears when communicating with different people in the social network. The avatars of different users may meet or interact in certain shared spaces in the virtual domain, and each user may want to control the way the avatar appears or react. This area calls for interdisciplinary research by engineers, psychologists, and sociologists on understanding and modeling the implications of the technology use in virtual spaces and the balancing factors influencing user's motivation, participation, and perceived control in social interactions. Pervasive social connectedness setups can then be developed to offer flexible control to a user on the choice of visual representation with each of the several people in concurrent interactions.
This domain of applications draws on the advent of flexible access to media content such as on-demand or search-based content delivery, as well as pervasive communication between multiple users to share experiences in a social network setting. The multimedia systems of the future merge various kinds of content such as movies, news, video phone calls, internet-based content, and will also offer context-based atmospheres with background music, color-controlled LED lighting, and projected patterns on walls based on the user's desired settings and acquired behavior model. This offers opportunities for research on the methods to learn the user choices, handle the concurrency of several services, as well as implications on the choices within a social setting. The ambient media system not only delivers content to the user, it can also observe the user, analyze the user's activities, learn about the user's preferences or lifestyle routines, and adapt its services to serve better. The system also acts as the information gateway between the user's home and the outside world, connecting the user based on the stated choices to others via pervasive communication links including video links and representation in virtual domains by avatars.
- Personal recommendations: Home and office wellness
The aim of this research area is to investigate how intelligent systems can employ interactions with users to obtain information that would be the most valuable additions to the knowledge base present in the system. Example applications could be deciding on the time and type of querying the user to help disambiguate an observation that the system has inadequate information to resolve, or designing a set of instructions for the user to follow during an interaction in order to learn certain activity or object in the environment or to discover the user's preference level on a certain service before the service is rendered to the user. Many other application examples exist which involve the notion of proactively designing interactions with the user to accumulate knowledge.
Within these application domains the following projects are under investigation at A I R Lab:
A I R Lab