Starting in September, the Soling group continued advancement of a project to develop a device which would enable quadriplegics to "touch type" and navigate the web. A group of REU students from several universities first began work on the project in the spring of 1996. Our efforts were concentrated toward further development of sensors and the mechanism used to orient the sensors and hold them in place. Eyal Sherman, who is a fifteen year old quadriplegic has been a part of this project since it began and worked with us in evaluating and testing design prototypes.

Early Designs

Our first task was to determine what had already been done. We reviewed taped sessions with Eyal and noted his facial movement capabilities. It was necessary to take maximum advantage of the muscles Eyal had available to him, to allow him to optimally control the activation of sensors interfaced with a program called Neattools developed by I3. The previous team's most recent design consisted of two photosensors that were mounted on the shell of a construction helmet. These tow sensors were contained within a length of lockline so that they could be positioned near Eyal's cheeks, and were activated when he covered the surface, blocking the light. A third sensor was placed underneath the chin on a metal rod, adjusted to a semi-elliptical shape so that it dropped down underneath the chin and could be attached by its ends to the outside surfaces of Eyal's head rest.

Goals and Constraints

The 'helmet' design allowed Eyal to successfully play both Tetris and Pacman. We used this design as well as previous designs to formulate goals and establish constraints for a new design. Our overall goal remained the same, to allow Eyal a means to actively interact with his environment, specifically to be able to type and control the computer. To attain this goal we defined the following objectives: to find a new type of sensor which Eyal could easily control with the motion he was capable of using his forehead muscles, to increase the reliability of the signals, to decrease the set up time, and to increase the adjustability of the 'sensor-holding' mechanisms. In addition to these objectives the constraints we applied to our design were that it be cost efficient, comfortable, unobtrusive to the user, not chair specific, and as inconspicuous as possible.

Our Design Process

We began our design process by considering possibilities for the forehead sensor because we wanted to make use of a fourth channel. The first sensor we tried was a Hall Effect transducer provided by Dr. Ed Lipson. The initial idea was to force the electrode against the surface of Eyal's forehead. Thus as Eyal raised his forehead, the electrode would slide closer to an attached magnet creating a signal. The problem with this was that we could not get the transducer to actually slide across his forehead consistently because we were not able to keep the pressure constant holding the transducer against his head.

The next idea for the forehead sensor involved the use of a wheel and a strain gauge. The wheel was extended down vertically along the midline of his forehead. Ideally, as Eyal lifted his eyebrows the wheel would rotate and the change in resistance in the attached gauge would trigger a signal. This idea was fine in theory, however when we tried it, we realized there was not enough friction between the wheel and the forehead to actually move the wheel consistently. The other problem was that when the wheel did in fact move, it would get stuck in place. Thus the signal would not be constant.

Two new sensors were also considered: microswitch and proximity sensor. The microswitch was discarded due to variability problems. The on/off nature of the switch was a problem for the software. It tended to flood adjacent channels. The proximity switch was abandoned because it was expensive. Research indicated it was an on/off switch as well.

The first idea for the photosensors was to eliminate the helmet. The alternative we considered was to use glasses to hold the sensors. Eyal wears glasses normally so we thought that glasses would be more comfortable, less obtrusive, and decrease set up time. We built a model from wire tubing, and mounted photosensors on the bottom edge of the frame. We intended for him to be able to move his cheek and establish a signal. In testing we realized that Eyal's maximum cheek motion, was not below his eyes, but it was on the side of his cheeks. We also encountered a problem because the sensors were not adjustable.

A second idea for replacing the helmet was to use the lining of a construction helmet, which could be placed directly on Eyal's head to support sensors. This was a problem, because the sensors could not be mounted rigidly on this structure. We experimented with mounting the Hall Effect transducer in this manner, and it was not effective.

The Current Design

After an unsuccessful session with Eyal, we needed to regroup and collaborate towards a different direction. The group concluded to construct a new mounting system. (See picture one) Two metal sleeves were made to fit over Eyal's headrest. Foam pieces were used to stabilize his head and add comfort. These pieces were held to the headrest with C-clamps. The sensors were mounted to these sleeves.

Extreme adjustability was needed for the cheek sensors. Picture two shows the apparatus used.

Photocells were extended from these sleeves with the use of extending antennae and adjusted with double ball joints. They were attached to the sleeves by a wheel from an erector set. This added another element of adjustability. For fine tune positioning of the photocells elbow hinge joints from plastic action figures were placed at the end of the antennae. Finally cones were placed around the photocells to reduce the effects of changes in ambient light which effected calibration of the cells. Picture Two

A new chin sensor was made consisting of a photosensor and hollow aluminum tubing. It was reduced to one arm extending from the sleeve similar to a microphone on a headset. It fashioned similar to the previous design. The tubing was connected to the sleeve using the same wheel mechanism as the cheek sensors. Some modification of this design is still needed.

The current forehead sensor was designed by Mark Bechtold, an industrial design student. It is a headband with two attached potentiometers which are connected by a wire that lies across the forehead, resting directly above the eyebrows. By raising his eyebrows, Eyal can change the resistance in the potentiometers creating a signal. Picture three shows Eyal using the forehead sensor.

This design proved to be the most effective to date. It allowed Eyal to use four channels to interact with the computer. Picture four documents Eyal playing Pacman through the program, Neattools.

Recommendation for the future

In the immediate future we have three immediate concerns which we plan to address. We would like to use light emitting diodes as a light source to eliminate the dependency on ambient light. We are considering using gel filters to help with this problem. Clamping of the sleeves is a second concern. Bands wrapping underneath the headrest could eliminate the use of C-clamps. Finally, the sleeve material should be changed to plastic. This will increase longevity, improve the appearance of the sleeves, and decrease wear of the headrest.

We have also considered long term possibilities. We feel that introducing feedback to the sensors would aid the user in knowing when the signals are sufficient. Fiber optics could be used as a replacement for ambient light or the light emitting diodes. The current problem with this is the high expense of both the fiber optics and a sufficient light source for coupling. Two different types of sensors could be explored further. To replace the Hall Effect transducer a differential variable reluctance transducer (DVRT) is a possibility. The team is currently pursuing this option.