Context-Aware Work Surfaces - Semantic Scholar

5 downloads 3746 Views 262KB Size Report
hospital, and task-aware lighting for a desk. ... sensing and flexible actuation systems to allow the context- .... faucet, a desk light or a drill press (see Figure 2).
CHI submission #568

Context-Aware Work Surfaces Leonardo Bonanni, Chia-Hsun Lee, Ted Selker MIT Media Laboratory 20 Ames Street Cambridge, MA 02139 +617 253 4564 {amerigo, jackylee, selker}@media.mit.edu ABSTRACT

for persons with disabilities.

Context-aware interfaces such as the computer desktop work to anticipate a user’s needs and desires in order to simplify tasks. We demonstrate that such principles can be applied to physical work surfaces to provide various types of hands-free assistance based on the needs of users and their tasks. Context-aware work surfaces use vision recognition in combination with a model of the user and the task to improve the accessibility, safety, comfort and efficiency of interfacing with these work surfaces. These work surfaces can support varied useful scenarios robustly. Their adaptable architecture can be applied to work surfaces as diverse as computer desks, kitchen countertops and hospital sinks. Context-Aware work surfaces operate based on user models that consider ergonomics and sensor placement, the expected task and environment, and recognize when these parameters are exceeded. These work surfaces automatically adjust their height and slope to individual users and provide task-specific assistance such as a hands-free tap for a sink, hand-washing compliance for hospital, and task-aware lighting for a desk. This paper describes the design of context-aware work surfaces, presents our working prototypes and discusses the results of user evaluations of our system.

INTRODUCTION

We are coming to expect that our digital desktop is contextaware and even anticipates our needs. What if our physical work surfaces could provide the same benefits? The costs and benefits of physical work surfaces are far more tangible than those of digital desktops. The physical world is hostile to both technology and people. Improperly designed work surfaces can lead to short- and long-term injury [9]. In many cases, a work surface that is suitable for able-bodied individuals precludes handicap accessibility and vice-versa [1]. Work surfaces that require manual contact with controls can lead to the spread of disease [4,5,6]. The same environments that pose dangers to humans can damage machines and reduce the effectiveness of sensors and actuators.

Author Keywords

Context-Aware Computing, Ergonomics, Human Factors Engineering, Interaction, Smart Rooms, Image Understanding, User Evaluation. Figure 1. Context-Aware Work Surfaces: SmartSink (left), SmartCounter(middle), SmartDesk(right).

ACM Classification Keywords

Categories and subject descriptors: H.4.m [Information Systems Applications]: Miscellaneous, Work Surfaces; J.7 [Computers in Other Systems]: Consumer products; H.5.2 [User Interfaces]: Ergonomics; K.4.1 [Public Policy Issues]: Human safety; K.4.2 [Social Issues]: Assistive technologies

While many types of digitally-augmented work surfaces have been proposed, the technologies they employ rarely translate to real work such as operating a drill press or chopping vegetables. We present robust, real-world applications of context-aware technology to make physical work surfaces safer, more accessible and more efficient to use. Rather than require specialized interfaces or artificial behavior, these work surfaces respond to users engaged in natural tasks in a work environment such as a kitchen, office, or hospital. The context-aware work surfaces seek to provide a variety of hands-free assistance without interfering with the user or distracting them from their task. 1

CHI submission #568 Our choice of vision-based sensing improves overall precision and adaptability compared with the simple infrared systems.

In addition to performing a pre-determined set of hands-free assistance, context-aware work surfaces can be easily updated to serve additional tasks. We chose vision-based sensing and flexible actuation systems to allow the contextaware work surfaces to benefit users in a variety of scenarios. Initially, the context-aware work surface was developed for use as a kitchen sink. By initially considering a heavy-duty application with the added factors of dirt and water, we hope to develop a system robust enough to be applied in almost any other context. The first application, SmartSink, is capable of adjusting automatically to a user’s height regardless of whether they are seated or standing (see Figure 1). In addition, SmartSink expands on the capability of traditional handsfree faucets to control both water flow and temperature based on an understanding of the task being performed in the sink. We next apply SmartSink to the problem of handwashing compliance in the context of hospitals. We then modify SmartSink to become SmartCounter, a work surface that adjusts height to provide the best leverage for various cooking-related tasks. Finally, we apply the technology behind SmartSink and SmartCounter to SmartDesk, a lightduty work surface that automatically adapts height, slope and lighting based on the user and task. A number of actuators can be added to context-aware work surfaces in order to make them more effective, including controls for surrounding lights and doors.

Another type of widely-used work surface automation is height- and slope-adjustment of tables in work places to provide an ergonomic work environment for employees. Such systems are recommended by the Occupational Safety and Health Administration to prevent repetitive stress injuries to manufacturing employees [9]. These adjustment systems require direct control and thus interrupt normal behavior. Direct-control height- and slope-adjustment systems are not suitable for tasks in which multiple users share the same work surfaces and are restricted from touching controls because their hands are occupied or they need to avoid contamination (i.e. commercial kitchens, hospitals). Foot pedals and lean-bars can also be used to control such work surfaces, but they require artificial, counterintuitive behavior that can still interfere with the task at hand. One kitchen of the future that allows users to adjust the counter height for different tasks found that direct-control motorized adjustment led to users performing the same tasks at different heights on different occasions[8]. This is likely because users do not know the proper ergonomics of work surfaces, and because typical adjustment systems depend on users to know exactly when to release an “up” or “down” button rather than dialing in a precise position. We adopt similar height- and slopeadjustment as these systems, while adding the element of automation to ensure that adjustment happens precisely and without interfering with normal user behavior.

User evaluations, regulatory guidelines and professional experts have confirmed that these context-aware work surfaces can have a direct impact on the safety, efficiency and accessibility of tasks performed on context-aware work surfaces. We plan to apply the context-aware model to work surfaces of greater complexity and develop further applications for the vision sensing and flexible actuation already in existence.

Digital augmentation of physical work surfaces has been demonstrated to expand the capability of traditional work in numerous research projects. The Digital Desk [15] augments paper-based work such as writing and drawing with digital benefits such as instant calculation, copying, saving, and spell-checking. The system can be quite useful for tasks that are also easy to perform on a computer, but it has no impact on real-world work. Various Tangible User Interfaces (TUIs) have been built to augment physical work surfaces to perform task-specific calculations through artificial, dedicated tools [7]. The metaDESK [12] projects information on a table top based on the manipulation of dedicated objects with fixed functions, such as an RFIDtagged magnifying glass to cue a digital zoom. The Sensetable [10] projects task-specific information such as network routing diagrams based on the physical manipulation of RFID-tagged and capacitive-sensing “pucks” with variable value on a table top with an overhanging projector. In both systems, the work surface can only be used for dedicated, artificial tasks that are not performed naturally elsewhere. The I/O bulb and Urp combine to form a work surface in which three-dimensional manipulation becomes possible, but only for representational tasks such as manipulating building models to observe their cast shadows projected by a multi-media projector [13]. Chameleon Tables are height-adjustable

RELATED WORK

Context-aware work surfaces arise from two distinct areas of research in intelligent environments: human factors engineering and computer-human interaction. A number of products and research projects have influenced our work and helped us to identify opportunities and requirements for context-aware work surfaces. Context-aware adaptations to work surfaces have been readily accepted into the market, most notably proximity-sensitive automatic faucets. These ingenious devices are pervasive in multi-user bathrooms for their benefits in reducing water use up to 70% and avoiding the need for physical contact with contaminated taps [2]. The same family of simple, robust context-aware fixtures for multi-user environments has been expanded to include automatic soap dispensers, air fresheners and toilet-cleaning systems. However, the infrared range-sensor technology on which these products are based can be unreliable and is limited to sensing proximity, without any ability to understanding context. Our context-aware work surfaces adopt the sensor placement of these systems by concentrating sensing on an overhanging projector arm.

2

CHI submission #568 sink’s faucet is a highly augmented version of proximitysensing automatic faucets. In addition to a proximity sensor that serves to initiate the system, the overhanging structure supports two webcams: one to observe the user, and another to observe the task. Initially, each camera corresponds to an independent system and correlates to a specific type of actuation: the task camera informs the flow and temperature of the water exiting the tap, while the user camera activates an automatic motorized height-adjustment system (See Figure 3). As the SmartSink gains functionality, the vision recognition is gradually enhanced while additional actuation is modularly added to the system. We have implemented six applications for sensing and actuation of context-aware work surfaces: automatic height adjustment, automatic water flow and temperature control, handwashing compliance, display, task lighting, and task-based ergonomic models.

café tables that are aware of their own height and change information displayed on their LCD screen surface based on whether they are being used as café or restaurant tables [11]. The combined benefit of self-awareness and task-awareness demonstrated in these research projects inspired ContextAware Work Surfaces with the added goal of augmenting real, physical task in a way that would provide tangible, physical benefit. DESIGN

Context-aware work surfaces provide diverse types of hands-free assistance through adaptable sensing and actuation based on a model of the user and task. We began with a study of existing work surfaces in residential, commercial and industrial environments. In each case, we observe formal and functional similarities that serve as the basis for the design of context-aware work surfaces. Work surfaces ranging from sinks to milling machines share the common morphology of a working area with an overhanging structure housing the tool, whether it is a faucet, a desk light or a drill press (see Figure 2).

Figure 2. Different Work Surfaces share the same morphology of a plane with an overhanging projector: kitchen sink (left), computer desk (middle), drill press (right).

Figure 3. SmartSink (left) and an input/output diagram of context-aware work surfaces.

The design of context-aware work surfaces seeks to achieve general principles for augmenting real-world task-oriented objects. We sought to avoided the need for sensing to be distributed throughout a space and instead attempted to find the simplest, most compact design with the greatest potential for adaptability to a wide range of tasks. Contextaware work surfaces do not require artificial behavior to adapt and respond according to a model of the user and task. This model considers three aspects of context: the ergonomics of the user, the activity being performed, and anything outside normal parameters that might endanger users. The feedback is designed to be unobtrusive so that any automation can not detract from the task being performed, and can only help.

Automatic Height Adjustment

The height of a work surface can have great impact on its safety, comfort and accessibility. Imagine, for example, a sink in a commercial kitchen where cooks of different stature empty pots of boiling water. If the sink is too low, a tall cook might splash boiling water outside the sink and cause severe burns to himself and others. If the sink is too high, a short cook might not be able to lift the pot over the edge of the sink – with equally catastrophic consequences. More commonly, work surfaces that are not at the right height for users cause ergonomic injuries, especially when used in a standing position. The Americans with Disabilities Act’s Uniform Federal Accessibility Standards specifies work surface heights for use by wheelchair-bound individuals [3]. The ADA mandates that at least one fixture in public places be accessible to wheelchair-bound individuals, directing that these bathrooms and kitchens have multiple sinks or that their only sink be too low for many standing users. The

IMPLEMENTATION

Our first prototype of a context-aware work surface seeks to tackle one of the most challenging of environments for technology: the kitchen. SmartSink [3] is the initial working prototype of a context-aware work surface. The

3

CHI submission #568 requirements for an accessible lavatory, for example, make for a very uncomfortable standing lavatory (See Figure 4).

the motorized sink up to the proper height with a stepper motor through a PIC-based microcomputer controller.

Figure 4. ADA lavatory clearances.

By comparing the height to a model derived from a combination of ergonomics information and handicap accessibility guidelines, the surface is lifted to rest a bit below the elbow height of the user (See Figure 6). The height of the work surface is designed to reach the elbow of most users within less than one second after they approach the sink frontally. Initially, the sink is at rest in its universally accessible configuration at 28” high. This is to prevent the sink from remaining in a high, inaccessible position in case of a power failure.

Along with the ADA, homebuilders are providing accessible bathrooms in anticipation of aging population. In a typical residential kitchen, the sink and countertop are not accessible because these work surfaces are unresponsive to the user. We see this project as a first step towards making accessibility and ergonomics the norm rather than the exception in residential as well as commercial and industrial interiors.

Figure 6. Our ergonomic model for determining work surface height based on user height. Automatic Water Flow and Temperature Adjustment

Imagine never having to touch a water faucet - it leaves your hands free to work with the materials you want to put under the stream of water: a large pot, a colander of vegetables, or your dirty hands. We designed a system to automatically turn water on to an appropriate temperature and flow when something is presented to the sink. A second webcam mounted to the faucet observes the contents of the sink and classifies them through color-recognition and shape-recognition algorithms in C++ using Microsoft Vision SDK libraries and running on a PC. When a user presents vegetables, the program interprets their green color and dispenses cold water for washing them. When a user presents their hands, the computer recognizes the color of flesh and dispenses warm water for washing hands. When a user presents pots and pans, the computer recognizes the black (iron) and white (glare) and dispenses hot water for washing pots. A user can also train the program through a neural network by specifying the object and the desired water temperature to a PC interface. A PIC-based microcomputer actuates two electric water valves in the tap and an instantaneous heater on the supply line to control water flow and temperature.

Figure 5. The lift mechanism (left) and two different work surface heights (right) with height tracker screenshots (inset).

Comfortable work surface heights vary extremely between tall standing users and children or handicapped/seated users. SmartSink is mounted on a track that allows a vertical adjustment from 71cm to 122cm high, suitable for users as tall as 203cm. A webcam mounted on the overhanging faucet observes the user in front of the sink. The height tracker is a C++ program utilizing the Microsoft Vision SDK Library to determine the user’s height. An infrared proximity sensor activates the height tracker once a user approaches. The height tracker continuously compares video frames from the camera to adjacent frames to detect a user’s motion (see Figure 5). When it establishes the location of highest (y-axis) movement, the program drives 4

CHI submission #568

Figure 8. Display inside the stream of water: hot (left) and cold (right). Figure 7. Object recognition of objects placed in the SmartSink: screenshot (left), Pot (upper right), Vegetables (lower right).

Task Lighting

After demonstrating the adaptability of SmartSink to a variety of heavy-duty institutional application, we sought to apply the system to light-duty residential and office work surfaces. While an automatic faucet is most useful in a sink application, the same overhanging sensor-enabled arm is suitable for providing context-aware task lighting in a desk or countertop application. As in the sink, the task camera recognizes the task being performed and responds with varied illumination of the work surface. In the case of a computer desk, the task lighting is bright for work on paper and dim for work on a laptop computer (see Figure 9). Further modifications include glare-free lighting to direct the light according to both the task and the user’s stature. Although we achieved the proof of concept by remotely activating spotlights in the ceiling of our lab, all of the above lighting requirements are easily achieved with an array of LEDs mounted to the overhanging arm. The array can be dimmed or turned on zone by zone to vary both the intensity and direction of task illumination.

Hand-washing Compliance

A recent review of 11 studies noted that the level of compliance with basic hand-washing protocol by hospital staff ranged from 16% to 81% [14]. Since dirty hands are a primary means of spreading infection, increasing handwashing compliance can have a direct positive impact on reducing the 80,000 deaths a year due to nosocomial infections contracted in US hospitals. SmartSink’s task camera was already enabled to recognize hands in order to turn on warm water for washing them. Without modifying the structure of the sink, we were able to include a software functionality that ensures that hands stay under the flow of warm water for at least 10 seconds. We considered various types of feedback to encourage compliance, including displays and acoustic feedback. Since these do not actually enforce hand-washing; we decided to implement a more robust means of enforcing hand-washing compliance: the sink is located outside the entrance to an examination room that remains dimly lit with a locked door. To open the door and turn the lights on, hospital staff wishing to enter must wash their hands for the full 10 seconds. Display

In addition to providing sink-specific assistance, SmartSink includes a downward-projecting display housed inside the faucet similar in function to the multimedia projectors used in many augmented work surfaces. Red and blue LEDs inside the faucet project colored light inside the stream of water to indicate the temperature of the water. This display communicates the context-aware system’s decision and eliminates the need to wet one’s hands in order to confirm water temperature when, for example, filling a heavy pot. In later implementations, these LEDs become task lighting that varies based on what the work surface is being used for. Unlike multimedia projectors, the illumination in context-aware work surfaces is minimally distracting and seeks to increase the visibility of the physical task.

Figure 9. Different task lighting for Smart Desk: paper-based work (left) and computer work (right). Task-Based Ergonomic Models

In addition to applying the SmartSink chassis to multiple types of tasks, we considered the variation in comfortable work surface height according to the task performed. Like the glare-free task lighting, this system requires communication between the task-recognition and userrecognition systems to work most effectively. In pilot studies, we found that a sink used for pouring out a heavy pot of boiling water needs to be as much as 20cm lower

5

CHI submission #568 than a countertop used for preparing food. This is due in part to the fact that work surfaces used for delicate operations should be as close as possible to a user’s eyes for maximum visibility, whereas operations requiring brute force often need to be far away to allow for proper leverage. When we use the same context-aware work surface for multiple uses by simply covering the sink with a board to turn it into a desk, the height tracker takes into account the status of the work surface through the task camera and applies the appropriate ergonomic model.

in written comments of two users and did not deter from their overall opinion of the system. We found the results of our questionnaires to be statistically significant in a paired samples t-test with p0.1). For the two questions “Was the counter at a comfortable height?” and “Do you feel that this counter would be useful?”, eight out of nine users preferred the context-aware work surface to the control surface. We expected that subjects would be surprised and even startled by the immediate movement of the counter as they approach, but this effect was only noted

A representative of a major hospital management company expressed interest in applying the sensing and design to the problem of hand-washing compliance. 6

CHI submission #568 Regulation

assistance to first-time users. By adapting the same system to a kitchen sink, a hospital sink, a countertop and a computer desk, we learned that the system is precise, robust and easily upgraded to take on additional functionality. The use of vision-based sensing improves the accuracy and adaptability of this context-aware appliance as compared to existing, infrared-based systems. Although complex, vision recognition is an excellent way to understand user and task in a constrained physical environment with a limited set of users. Most importantly, the sophisticated sensing can be adapted through software upgrades to a variety of tasks from height tracking to face recognition. Likewise, the modular actuation scheme allows the system to have a measurable impact on health and safety as demonstrated by the hand-washing compliance appliance. The application of sophisticated computer-human interaction to real-world work surfaces can have a significant impact on the safety, comfort and efficiency of the tasks performed at these surfaces.

The greatest motivation for developing context-aware work surfaces stems from the combination of OSHA, ADA, FDA and CDC regulations that together encourage ergonomic, accessible, and sanitary design of work surfaces in places ranging from day-care centers to heavy manufacturing plants. Context-aware work surfaces demonstrate that a single surface can provide safe, comfortable and efficient work conditions for a variety of individuals involved in diverse tasks. Height adjustment and task lighting reduce physical strain, improve visibility and reduce the risk of injury. Hands-free actuation of water, light, and doors limits the risk of spreading infection while hand-washing compliance reduces the bacterial load of contaminated hands. A single, robustly-designed chassis can offer these benefits to a variety of existing and future work surfaces in any environment. Perhaps most importantly, these contextaware work surfaces demonstrate how the integration of digital technology is necessary to make a single physical appliance capable of satisfying the requirements of any users in a multitude of contexts.

REFERENCES

1.Americans with Disabilities Act http://www.usdoj.gov/crt/ada/adahom1.htm

FUTURE WORK

Context-aware work surfaces are a platform for future adaptation into various types of tasks. Cooking, cleaning, and desk work have already been demonstrated to benefit from context-awareness. The next step is to implement further advancements and install them into active environments for long-term evaluation and ultimate implementation. The task-oriented camera can evolve to recognize a variety of activities, including the use of specialized tools such as in machining. A context-aware shop tool such as a table saw or drill press could greatly reduce risk of injury by shutting off moving parts when users are not properly positioned, or if they forget to wear safety goggles. Ventilation and acoustics can also be modulated by the task-oriented camera, so that soldering and concentrated desk work can happen in the same place without requiring manual adjustment.

2.Automatic Sensor Faucets http://www.airdelights.com/faucet.html 3. Bonanni, L., C.H. Lee, S. Sarcia, “SmartSink: Context-

Aware Work Surface.” In CD-ROM SIGGRAPH 2004. 4.Centers for Disease Control www.cdc.gov 5.Food and Drug Administration Food Code http://www.cfsan.fda.gov/~dms/fc01-toc.html 6.International Scientific Forum on Home Hygiene, “Guidelines for Prevention of Infection and Cross Infection in the Domestic Environment.” Milano, Italy: Intramed Communications s.r.l., 1998. 7.Ishii, H. and Ullmer, B., “Tangible Bits: Towards Seamless Interfaces between People, Bits and Atoms,” in Proc. Conference on Human Factors in Computing Systems CHI ’97, ACM Press, pp. 234-241, 1997.

The ultimate benefit of context-aware work surfaces is the elimination of a need for multiple, fixed appliances to satisfy both standing and seated users. Further development of this principle can lead to environments that are capable of being used for an ever-increasing variety of tasks with less and less physical infrastructure. Digital technology makes it possible for the same physical surface to serve a variety of tasks never before possible.

8.Mori, T. Kuroiwa, T. Morishita, H. Sato, T. “Assistance with Human Actions and Individuality Adaptation by Robotic Kitchen Counter.” In Proc. ASER ’04. 9.Occupational Safety & Health Administration www.osha.gov 10.Patten, J., Ishii, H., Hines, J., Pangaro, G., “Sensetable: A Wireless Object Tracking Platform for Tangible User Interfaces.” In Proc. CHI ’01, ACM Press (2001).

CONCLUSION

Context-aware work surfaces are real-world systems that seek to provide hands-free assistance to any user based on an understanding of the task and the user. The system is based on a work area with an overhanging input/output fixture that serves to recognize the user and task as well as to provide task lighting and water in the case of a faucet. User evaluation has demonstrated that despite their novelty, context-aware work surfaces provide comfortable, useful

11.Selker, T., Arroyo E., Burleson, W. “Chameleon Tables: Using Context Information in Everyday Objects.” In Proc. CHI ’02. 12.Ullmer, B. and Ishii, H., “The metaDESK: Models and Prototypes for Tangible User Interfaces.” In Proc. UIST ’97.

7

CHI submission #568 13.Underkoffler, J., Ishii, H., “Urp: A Luminous-Tangible Workbench for Urban planning and Design.” In Proc. CHI ’99.

Practices.” Agency for Healthcare Research and Quality, Contract No. 290-97-0013. http://www.ahrq.gov/clinic/ptsafety/

14.University of California at San Francisco – Stanford University Evidence-based Practice Center. “Making Health Care Safer: A Critical Analysis of Patient Safety

15.Wellner, P., “Interacting with Paper on the Digital Desk,” Communications of the ACM, 36(7), 86,96, 1993.

8