2. This paper describes a sensory system for
implementing a human–computer interface based on
electrooculography. An acquisition system captures
electrooculograms and transmits them through the
ZigBee protocol. The data acquired are analysed in
real time using a microcontroller-based platform
running the Linux operating system. The continuous
wavelet transform and neural network are used to
process and analyse the signals to obtain highly
reliable results in real time. To enhance system
usability, the graphical interface is projected onto
special eyewear, which is also used to position the
signal-capturing electrodes.

3. The purpose of this research paper is to develop a system to capture and
analyse EOG signals in order to implement an HCI, as shown in Figure 1.
The system comprises two electronic modules—the signal Acquisition
Module (AM) and the Processing Module (PM). Eyewear incorporating a
set of appropriately positioned dry electrodes captures the EOG
signals, which the AM acquires, digitizes and transmits using the ZigBee
protocol. The PM receives the signals from the AM and executes the
algorithms to detect the direction of the user’s gaze. Simultaneously, it
projects the user interface onto the eyewear and, according to the
selection made by the user, transmits the commands via WiFi to a home
automation system or performs other tasks (i.e., call a nurse, etc.).

4. The electrooculogram captured by five electrodes placed around
the eyes. The EOG signals are obtained by placing two electrodes
right and left (A-B) to detect horizontal movement and another
pair above and below the left eye (C-D) to detect vertical
movement. A reference electrode is placed above the right eye (E).
The eyewear has a composite video input (PAL format) and
displays high-colour, high-contrast images at 320 × 240
resolution, equivalent to a 46-inch screen viewed at a distance of 3
metres.

6. Analog signal acquisition hardware includes two differential input (CH1, CH2), which
are digitized internal ADC of the microcontroller (LPC1756, 12-bit
resolution, sampling at 100-300 Hz frequency, ridden in steps of 10 Hz) before being
transferred via the protocol ZigBee.
Two channel amplifier was designed to get bioelectric signals, each channel can be
configured dynamically and individually (the active adjustment of the channel, the
channel offset, sampling frequency or amplification circuit) via commands
transmitted via the protocol ZigBee.

7. Function Processing Module must receive signals from
EOG through the protocol ZigBee, apply the
appropriate algorithms to detect movement of the
user's eye, display the user interface and sends the
appropriate command through WiFi to your home
automation system for high-performance.
Processing Module based on SoC (system on
chip), OMAP3530, which includes a the kernel of the
cortex-A8 as well as the C64x + DSP, reaching 720
MHz. It has 512 MB ​of RAM and 512 MB ​of flash
memory. This provides a direct composite video output
(compatible with both PAL and NTSC
formats), coupled with the wrapper 230 Vuzix eyewear

8. Figure 5 shows the processing performed on the digitized
signal, EOG. processing consists of correction of signal
receiving, applying a linear model and intermittent eye trained
neural network according to signals from the user. Acquisition
module allows for adjustment of channel gain. Eye model
calculates the ratio between the change in EOG and eye
movements, as well as the calculation of the minimum threshold
of detection.

9. Detector blinking eyes determines consistently 2 or 3
blinks an eye when it is discarded interrupted eye
movements can be seen in figures. Block eye
movement detector determines the legality of
intermittent motion detection.

10. EOG signals are selected by 100 times per second. 1.68 ms are needed to
handle the CWT, while interrupted linear model takes 0.012 ms to detect
movement and determine the quantity of it. Blink detection takes 0.26
ms. A delay of 250 ms is required after the stick-slip is detected. Signal
propagation on RBF takes 8.52 ms. Finally, the block of the eye
movement detector requires 0.035 ms