Caleb Ollech & Chen Zhao
The OcuNav™ electro-oculographic peripheral controller is a PIC-controlled electronic device allowing users with paralyzed or limited-mobility to control USB-connected devices in their immediate environment using only shifts of gaze and winks of the right and left eyes. This technology leverages the polarity of the human eyeball by monitoring the voltage differential created between the left and right temples as gaze shifts from left to right (or vice versa), and using fluctuations in this differential as a natural switch to turn on/off simple household devices.
BACKGROUND & SIGNIFICANCE:
The biomedical community has long made efforts to develop assistive technologies for those with limited mobility and motor deficiencies. Subpopulations affected by motor neuron disease and quadriplegia, which account for over 125,000 people in the United States as of 2011, are two groups in which this need is apparent. Many eye-tracking or single switch-driven devices have been developed to address the communicative needs of these populations, though little commercial effort has been put towards allowing persons with limited mobility to retain direct control of devices in their immediate vicinity.
Environmental control systems, as these devices are formally named, are a class of electronic assistive technology enabling people with significant disabilities to independently access and operate equipment in their environment (e.g. home or hospital). Technology that minimizes a disabled person’s dependence on supplementary care is important not only to the patient, but also for his or her caretaker. Robust and effective environmental control systems relieve caretakers of trivial tasks such as turning on/off or adjusting the television, lights, etcetera, which allows for more attention to be paid to areas of greater clinical importance. Most devices in this category are reliant on a single switch, though some models are capable of responding to more than one user input type. Existing camera and computer-driven eye-tracking systems offer reliable navigation with high rates of speed and accuracy, but these devices come at substantial cost to the user. Additionally, many devices that have become standards in the realm of environmental control systems, such as the TASH Relax single-switch line of infrared remote control devices, are rudimentary by today’s technological standards and still somewhat expensive. This being the case, we believe there is a need for a novel, inexpensive, and reliable control system for persons with diabilities.
The OcuNav™ controller operates on the principle of electrooculography, in which the electrical potential of the retina is monitored using electrodes placed on the right and left temples. Since the eye is naturally polar (positive near the pupil and negative near the rear), lateral eye movements such as gazes and winks register spikes in the voltage differential between the left/right eye and a grounded reference location (in our case, the bridge of the nose). By measuring the amplitude and pulse width of these spikes, eye movements can be used as a natural switch by which electronic peripherals can be toggled on/off.
As with many assistive biomedical devices, the OcuNav™ controller is comprised of both software and hardware components:
The OcuNav™ device is composed of (a) signal acquisition goggles, (b) a PIC18F467K22 microcontroller, (c) two custom-printed circuit boards (PCBs), and (d) peripheral USB devices.
The signal acquisition goggles are a modified pair of 3M safety glasses. Padded electrode mounts were installed on the nose bridge of the goggles, as well as inside each temple. This placement allowed for reliable signal acquisition of the voltage differential between the right and left eyes as compared to the neutral reference voltage in the center (ground). Bioelectric impulses registered by the electrodes were transmitted to a signal processing circuit via alligator clips soldered to coaxial cable embedded within the goggle frame. Shortly outside the arms of the glasses (where stems contacts the ears), coaxial cables routed from the left and right temples were capped with BNC connectors, allowing the goggles to be unplugged from the rest of the OcuNav™ device as a small, lightweight unit.
Two printed circuit boards were implemented in our device, arranged on top each other within the device housing to minimize the total footprint of the electronics. The top board, mounted flush with the top of the acrylic housing, consists of an LCD screen, three blue selector LEDs (indicating which device is selected by the user), three blue toggle LEDs (indicating whether a device is powered on/off), one green ready-indicator LED (powers on only when the bioelectric signal has returned to the 2.5V baseline and is ready for user input), one calibration LED (indicating the user has initiated the calibration sequence), as well as a push button for calibration which triggers Calibration Mode. All status LEDS are connected in series with USB peripherals, serving as direct indicators of whether or not the outputs are successfully toggled by the user.
The signal output by the circuitry described above is sampled and processed in the PIC18F467K22 microcontroller. The software used to program this microcontroller in the OcuNav™ device was coded in the C# language and functions in a sequential logic progression split into three loops or “modes” (Navigation Mode, Calibration Mode, Standby Mode) that are each triggered by a distinct user input (logic progression can be seen in Figure 4 of the Appendix). Each time the device is powered on, the software initiates a “System Initialization” sequence that configures the timers, port configurations, LCD screen, analog-to-digital converter (ADC), and capture/compare module (CCP) to settings that allow our software to run optimally (such configurations involve switching pins from inputs to outputs, changing ports from analog to digital, etc.).
In all modes, voltage sampling of processed bioelectric signal is accomplished via analog-to-digital (ADC) conversion, whereby the analog signal received from the signal acquisition goggles is discretely sampled at 10 ms intervals (100 Hz) and scaled to a range of 0–5V via a conversion factor in software. We found that the vast majority of ocular signal impulses peak within this range, but that occasionally a strong wink would deviate from the 2.5V baseline and rail out at the 0V lower limit or 5V upper limit (depending on whether it was a left or right wink). This did not appear to affect detection in any way.
Navigation Mode allows for switching between and on/off toggling of the LED lights and peripheral devices. Within the software code, the 0–5V scale described above was split into four “activation zones”, each separated by a 100 mV. In order to initiate a response from the OcuNav™ device, the signal modulated by user eye movements was required to stay within the upper and lower bounds of a given “activation zone” for a predetermined pulse width (150 ms for winks and 250 ms for gazes, as of our latest code version). Pulse width monitoring was accomplished via the capture-compare module, which compares the signal voltage level to those specified for each “activation zone” and allows our software to detect when a signal is or is not above a given threshold. When a sample is determined to be within a voltage range assigned to a device response, the software increments a counter and samples the voltage level again. By repeating this process for each 10 ms sampling interval, we are able to determine the duration that an oculographic signal lies within a given threshold range (sampling interval multiplied by consecutive samples in range). Applying this concept to our device’s navigational scheme, we assign pins on the PIC microcontroller to fire high or low in response to signal amplitude and pulse width, resulting in control of current flow through the indicator LEDs and peripheral devices.
Calibration Mode is activated via a push-button mounted on the device housing, which initiates a sequence of on-screen prompts that assists the user in setting personalized voltage maximums/minimums for each of the four ocular commands used with the OcuNav™ device. When the calibration sequence is called, the program initiates a sampling period of indeterminate length and directs the user to gaze left. After this action has been completed, the user (or caretaker) should press the button again, which will end the sampling period. This process is repeated for the right gaze, as well as winks of the left and right eyes. During each sampling period, software tracks the maximum voltage value obtained and resets the corresponding “activation zones” to be centered around these new peak values more suited to the user.
Standby Mode is a feature of the OcuNav™ software that allows the user to indefinitely pause the device’s navigational capabilities with a wink of his or her right eye. This feature was implemented so that the user could suspend control of the device and be free to look around without inadvertently toggling peripherals on/off.
RESULTS & DISCUSSION:
Extensive testing of the OcuNav™ switching device was conducted in an effort to establish the most suitable default (non-calibrated) voltage thresholds and sensitivity levels (bounded voltage limits through which an ocular signal must pass in order to successfully trigger an action from the OcuNav™ device), as well as to determine functional specifications regarding the device’s receiver operating characteristic, quantify the incremental benefit of user calibration versus the default settings, and measure the device’s overall response accuracy in simulated real-world use.
Default Voltage Origin Optimization & ROC Analysis:
To best optimize our default voltage threshold settings for a wide user base, the voltage peaks inherent to all four ocular motions used as control commands in our device (right gaze, left gaze, right wink, and left wink) were quantified five times for three separate users using an oscilloscope. These peak voltage values were averaged across all three users (15 trials in total) to determine the optimal origins on which to center the threshold ranges for each ocular command. Averages of three sets of user data yielded mean voltage peaks of 3.17V for left gaze, 1.97V for right gaze, 4.99V for right wink, and 0.23V for left wink. Given the reproducibility of similar peak maximums and minimums through our testing, we opted to adopt these average figures as default voltage origins from which to expand the target window for each ocular command motion. (Testing process data can be seen in Figure 1 of the Appendix.)
Having optimized default voltage origins from which to expand the threshold ranges of each “activation zone”, we then went about finding the optimal voltage spread around these origin points that would yield the best combination of sensitivity and specificity when the device is used repeatedly. To accomplish this, we conducted ten trials of a test sequence that encompassed all possible device responses for six different size ranges of the “activation zone”. The ROC curve and voltage ranges tested can be seen in Figure 2 of the Appendix.
Overall Accuracy and Calibration Benefit Analysis:
Finally, we tested the benefit of the OcuNav™ device’s Calibration Mode by conducting 20 trials of the same test sequence described above (right gaze x2, left gaze x2, right wink x2, left wink x2, no action for 10 s) on three different users using the default voltages ranges and our own calibrated ranges.
User testing on the OcuNav™ controller revealed that the device is substantially more accurate that we initially predicted. Using the default threshold settings, our three test subjects registered correct response rates of 96.11%, 92.78%, and 92.78% respectively. When Calibration Mode was used to adjust the parameters of the thresholded zones to each test subject, the correct response rate jumped 3-5% to 99.11%, 95.56%, and 97.78% respectively (Figure 3 of the Appendix).
Total current draw of the OcuNav™ device is 884mA (4.42W), or approximately $0.29 per day. The materials cost for the project was $139.15 (detailed bill of materials shown in Figure 7 of the Appendix).
To summarize, we feel that our final device is successful in providing mobility-limited patients with an independent, robust, and cost-effective method by which to control devices in their immediate surroundings. We feel there is potential need for such a device in the market, and that the concepts utilized in our design merit some consideration for use on a broader scale.
We would like to thank Dr. Patrick Wolf, Thomas Jochum, and Nick Bottenus of the Duke University Department of Biomedical Engineering for their support. We would also like to thank Manjiri Kshirsagar, Andrew Norwood, and Sagar Naik for their involvement in the design and construction of the OcuNav™ device.
APPENDIX (Data & Charts referenced above):