PROCEEDIGS OF THE 2006 ITERATIOAL COFERECE O ARTIFICIAL ITELLIGECE

Volume II Editor Hamid R. Arabnia

Associate Editors Oscar Castillo, David de la Fuente Georgios A.. Demetriou, David Dodds Peter M. LaMonica, Raymond A. Liuzzi Patricia Melin, Jose A. Olivas, Gene Simmons

Las Vegas, Nevada, USA June 26-29, 2006 © CSREA Press

Conference on Artificial Intelligence |ICAI’06|

439

Intelligent Virtual Companion System for Independent Living Michael Todd Sreela Sasi, Ph.D. Gannon University Erie, PA 16541 [email protected], [email protected]

Abstract As people age, physical and mental abilities begin to degenerate. This often leads to disabilities. Many family members may not always be able to help the elderly, and home medical care can be discouragingly expensive. Currently, there are existing agencies that provide companionship services for the elderly by sending hired staff to the homes of elderly people who choose to live independently. The main goal of independent living facilities is to maximize the independence and productivity of individuals with disabilities, and their integration and inclusiveness into mainstream society. A component of independent living is an elderly companionship service. In this research, an Intelligent Virtual Companion System for Independent Living (IVCSIL) is designed to add a social companionship aspect to smart home systems. This artificial intelligent system will initiate casual conversation, provide cues for performing daily living tasks, monitoring a subject’s health, and perform secretarial tasks which will enhance autonomy.

Keywords: Intelligent Smart Home, Assisted Living, Virtual Companion, Artificial Intelligence, Independent Living

I. Introduction Independent living is more than living on one's own; its requirements can be complex and have many components. According to the Oregon Advocacy Center, “The philosophy of the independent living movement includes the values of choice and equal access [1]”. It’s goal is to maximize the independence and productivity of individuals with disabilities, and their integration and inclusion into the mainstream of American society.” A component of independent living is an elderly companionship service. As people age, physical

and mental abilities begin to degenerate. This can and often does lead to disabilities. According to the disability model [2] the encroachment of some reduced mental and physical abilities is a normal aspect of becoming older, and rejects the notion that persons with disabilities are in some inherent way "defective" and can not live on their own. The main area of disabilities is categorized to include: vision, mobility, hearing, speech and cognitive disabilities [3]. Therefore elderly people are excellent candidates for living in an independent living system, provided that the person with the disability is someone without a major handicap or impairment. Many elderly people do not want to go to nursing homes just because they have trouble doing everyday tasks at home [4]. Unfortunately, family members may not always be available to help, and home medical care can be discouragingly expensive for them. A solution is to develop and link a series of technical applications that will help elderly persons live safely, independently and cost effectively at home. According to a study in Queensland Australia, “Elderly people will be supported in their daily life by context aware homes, allowing them to age in their own home or familiar environment [5].” A context aware system will monitor the state of the elderly occupants. This will free nursing staff from the task of constantly supervising cursory needs of each elderly person. This will allow them more time to care about those who actually need their support most. Currently there are agencies that provide companionship services for the elderly. They run their services by sending hired staff to work at the homes of the elderly person on demand at

Conference on Artificial Intelligence |ICAI’06|

hourly costs. In the long run this can become too expensive to afford. In this research, an intelligent system is proposed for providing virtual social interaction and companionship support. Currently, a person’s social and companionship needs are not addressed in software designs for smart homes. The proposed research is intended to design a smart system that will provide elderly people with a diverse source of technical assistance and social companionship. The populations considered for this research are middle to lower class people, specifically the elderly people who intend to live independently.

II. Background The United States is currently experiencing and expecting a rapid increase in the number of its elderly people. The U.S. Census estimates that there will be about 12 million people over the age of 85 by 2040 who are expected to need physical and cognitive assistance. The greatest need for assistance will come from elders who live independently [6]. There are small independent companies who foresee a demand for independent living homes. The most well known and established service providers are: Elite Care, INTEGER, and CABA. Assistive living robotic products are not designed to consider the social, aesthetic, and emotional relationship issues. Current products focus too much on the client’s disability instead of enhancing the client’s quality of life [6]. The home is a central place for the aging population. The home environment has many components that can be enhanced by technology to assist elderly for independent living. In developing home features the approach is to take existing devices, and progressively add intelligence and build add-ons [7]. An example of the smart home development can be seen in the MavHome project. The project implements an intelligent agent to monitor the home with sensors acting on the home environment through device controllers [8]. Another system is the CARE system. It is a system that serves the audience in two ways. For the resident, it provides

440

biofeedback and cues to prolong their independence. For the monitoring staff, it spots the residents health problems early and reviews the quality control measurements [9]. In addition, a company called Freedom Scientific introduced a text to speech reader called, SARA (Scanning And Reading Appliance). It is an easy to use solution for reading a wide variety of printed text material such as books, mail, newspapers and magazines to visually impaired people. The SARA product uses optical character recognition technology to scan text. The scanned text can then be read aloud in a crisp, clear speech through speakers. The SARA device stores and remembers the contents of thousands of scanned pages [10]. This will make a powerful component to a smart home. Ultra Hal Assistant is a software which has built-in artificial intelligence. It performs as a PC desktop digital secretary and companion. It receives information from a user via text or speech. It can run applications such as ‘check email’ and ‘news’ from the internet. Its key feature is its ability to simulate human conversations. Conversations can grow with the user as the user feeds it more data about him or her. According to Thompson, smart devices and soft controllers must contain some capacity of evolution and some resemblance of intelligence [11]. Research on smart homes have been in the direction of applying the principles of ubiquitous, or omni present, computing [12]. These smart homes, though not commonplace, do exist. They work by adjusting itself to the residents’ needs. Their needs are defined according to the information provided and fed into the system. They should be able to communicate in an intuitive manner. Communicating directly with a context-aware environment enables them to achieve their goals more easily, and freeing their minds to think even further ahead of their current tasks and problems. A person naturally develops a predictable routine when in a familiar environment. Noticeable patterns can be found around focal points or centers of interest. In a home, patterns are found everywhere. For a smart home system to be available to the resident, the system must be accessible from

Conference on Artificial Intelligence |ICAI’06|

441

everywhere. Research shows that speech interface is the best way for smart systems to interact with the resident since it does not hinder household tasks, and speech can be done in most locations of the home [12]. By using existing equipments or by integrating them, an efficient system can be built that can provide social companionship for the elderly.

III. Intelligent Visual Companion System for Independent Living (IVCSIL) To accommodate the social and companionship needs in a smart home or in independent living housing, an Intelligent Visual Companion System for Independent Living (IVCSIL) is designed and being developed in this research. An architecture for IVCSIL system is shown in Figure 1. IVCSIL

IVCSIL

Communication Assistance

Health Reminder

Medicine Reminder

Diatary Suggestion

Secretarial Assistance

News with Internet Suport

Entertainment Assistance

Small Talk

Figure 1: Architecture for IVCSIL William Green, Diane Gyi, Roy Kalawsky, and David Atkins proposed that in order to get a proper survey on what the smart home should have, they used focus brainstorming, scenario building, group discussions, and design exercises with their sample test audience. They found that the needs, attitudes and expectations towards smart homes are not expected to be homogeneous across the population [13]. There are numerous factors influencing the views of the resident, which include sex, age, occupation, technical knowledge, wealth, occupancy of the home, lifestyle and culture. In order to have the companion system integrate to the wants and needs of the resident, a preliminary survey is conducted. This survey will limit and focus the decisions made by the IVCSIL. A sample survey is shown in Table1.

The preliminary survey was conducted among elderly people living in independent living facilities in Erie, Pennsylvania. In this survey, a person can also request additional features to be programmed into the system. The survey asks the elderly person if the lists of provided tasks are in fact the tasks they perform on a normal basis, and the duration of time it takes to perform the tasks. Each relevant task will be prioritized by the elderly person. If there are other major tasks that were not offered in the in the survey, the elderly person is free to add it in the list, and it will be handled accordingly in the same way as the other provided tasks. A database is constructed from the details provided by the elderly people.

Conference on Artificial Intelligence |ICAI’06|

442

Table 1: Sample survey Interview ee ID: Sex Living Facility: Date of survey: Interviewee Age: Time Frame Task From To AM/PM 7:00 9:00 AM Eat Breakfast AM Read News AM socialization 9:00 10:00 11:00 12:00

10:00 11:00 12:00 1:00

1:00 2:00 2:00 3:00 5:00 7:00

7:00 8:00

8:00 9:00

AM AM AM PM PM

Walk socialization read Eat Lunch socialization

PM PM PM PM

Walk socialization Eat Dinner socialization

PM PM

Walk socialization Read Leisurely

A1 Female Breviller Village, Erie Pa January 5, 2006 between 55 - 65 Task Task Notes/ Comments Duration Priority 2 hours 30min 30 to 40 minutes 30min variable 2 hours 30 to 40 minutes 30min

30 to 40 minutes 30min

1 3 2

eat talk read speed reader just small talk

1 1 4 1 2

view of the local lake talk about the walk just a time killer eat talk read just small talk

1 1 1 2

view of the local lake talk about the walk

1 1

view of the local lake talk about the walk

just small talk

9:00 10:00 PM 3 11:00 PM 7:00 AM sleep all night 1 Additional Information: She does not watch TV. Getting out is difficult since she does not have a car. She prefers to sing as her source of joy. She loves to read whenever she can. She does not take medicine regularly She sees a doctor about twice a year. She loves to talk about the word and eager to learn about everything.

An artificial intelligent system is designed based on if-then rules. The functionalities are shown in Figure 2 using Use Cases. The caretaker enters a detailed schedule for a person as text, and this is communicated to the end user using a text to voice interface. A task manager is designed in Visual Basic 6 where task priorities are identified and assigned. Conversation initiation is embedded in the scheduler. All other functionalities including secretarial, health, entertainment, and general living cues are embedded as subsystems. The conditions attached to the task calls on the individual functionalities will determine the next course of action of the system and expected responses from the user. The tasks are added through keyboard, and the system

while operational converts these texts into voice. For example, a functionality involving diet schedule will include the following: •

•

The system should remind a person 15 minutes before the meal to start to perform the meal task. When each meal is announced to the user, the system expects a response from the user. These expected responses are initialized in the settings of the database. The actual responses are then logged in a response database for further decision making.

Conference on Artificial Intelligence |ICAI’06|

•

•

•

Between each task, the system refers to its conversational functionality. The conversation will occur over a certain percentage of time between tasks on a particular topic. Within certain intervals of time without conversation, the system will initiate small talk from a list of topics in its conversation database. This occurs while the user is in the home, awake, and willing to talk. The system will not impose conversation or any other task on a user.

The functionalities of IVCSIL system is given in Figure 2 using use case diagram. The IVSCIL is designed and being developed with some basic functionality such as meals and health that has a successfully integrated social companionship component in it.

443

identified with a name, priority level, time in which to call the task and comments about that task. This system is designed to perpetually compare execution start time to the current time of day clock set on a PC. When a match exists, then a voice message appears noting that a task has been called at the prescribed time. Thus, the system will evolve as new tasks are added. The subsystem “greetings” is developed to generate messages including ‘good morning’, ‘how are you’, ‘did you sleep well’, ‘good night’ etc. The “small talk” subsystem will include ‘It is a little chilly, would you like more heat?’, ‘I noticed you are not saying much, is everything ok?’, ‘Your favorite athlete did well in a recent game’ etc. The “entertainment” subsystem will include ‘turn on your TV, your favorite show is about to start soon’, ‘Do you want to record a TV show in the near future?’ etc.

IV. Simulation The IVCSIL is a large system composed of many subsystems. In order to show that the main system is a working possibility, design and test implementation are done on the subsystems to show that the system components work. The subsystem focused on in this research is the user customizable meals and health focused task cues database. The system is developed using Visual Basic 6 combined with MSAccess. It is designed in a manner where navigation through each of its editing controls is command button based. When the system is activated it loads an existing database built of health and meals related tasks. The system provides an option for creating a personal database for the end user. In either case, the database is customizable. A user can scroll through the existing tasks. The user can edit or delete the desired task. New tasks can also be added. Each task is

V. Conclusion In this paper companionship is identified as a social concern dealing with the elderly in the American culture. The elderly currently have the option to use assistive living facilities or assisted living home care assistants. The current solution is costly and has service availability limitation. By expanding assisted living technology to encompass artificial intelligence (AI), a new helpful service can be offered that was not before. This service not only provides help in performing daily task and connecting with the outside world, it also has therapeutic and diversion aspect of simulated casual conversation. The result of the research concludes that the social dilemma can be alleviated in some extent through the use of AI technology.

Conference on Artificial Intelligence |ICAI’06|

444

Figure 2: IVCSIL UML Use Case

Computer Science, Millersville University, Millersville, PA, pp. 47.

VI. References [1] OAC, “Getting Assistive Technology for Independent Living,” Mar. 2005; http://www.oradvocacy.org/pubs/indlivingat .htm [2] D. Kaplan, “The Definition of Disability,” Mar. 2005; http://accessiblesociety.org/topics/ demographics-identity/dkaplanpaper.htm [3] B.W. Liffick, “Assistive Technology in Computer Science,” Department of

[4] NPR interview with Home Instead Senior Care co-founder Paul Hogan. http://www.npr.org/dmg/audioplayer.php?pr gCode=ME&showDate=24-Jul2003&segNum=8 [5] A. Rakotonirainy, S. Meyer, “A Survey of Research on Context-Aware Homes,” Australian Computer Society, 2003, pp. 2. [6] J. Forlizzi, “Robotic Products to Assist the Aging Population,” Interactions, vol. 12, Mar. +Apr. 2005, pp. 16-18.

Conference on Artificial Intelligence |ICAI’06|

445

[7] D.A. Norman, “Robots in the Home: What Might They Do?,” Interactions, vol. 12, Mar. +Apr. 2005, pp. 65.

www.freedomscientific.com/fs_products/sca nners_SARA.asp

[8] D.J. Cook, et al., “MavHome: An AgentBased Smart Home,” Proc. 1st IEEE Int’l Conf. Pervasive Computing and Comm. (PerCom’03). IEEE Computer Society, 2003. [9] Elite Care, “Creating an Autonomy-Risk Equilibrium™ (CARE),” Jan. 2005; http:// www.elite-care.com/oatfield-tech.html [10] Freedom Scientific, “SARA Scanning and Reading Appliance,” Mar. 2005; http://

[11] C.W. Thompson, “Smart Devices and Soft Controllers,” IEEE Internet Computing, Jan.+Feb. 2005, pp. 82-85. [12] T. Koskela, K. Väänänen-VainioMattila, “Evolution towards smart home environments: empirical evaluation of three user interfaces,” Pers Ubiquit Compute, Aug. 2004, pp. 234-240. [13] W. Green, et al., “Capturing user requirements for an integrated home environment,” ACM (NordiCHI ’04), Oct. 23-27, pp.255 – 258.