A wearable EMG and EOG device can turn facial muscle activity and eye movements into electrical signals for human-machine interfaces (HMIs). For automation engineers and system integrators, the design offers a practical example of how biopotential signals can be collected, conditioned, and prepared for embedded-system processing. Its central challenges are familiar to instrumentation work: low-amplitude signals, electrical interference, electrode placement, and the need to match an analog output to downstream electronics.
Why acquire facial EMG and EOG?
Electromyography (EMG) measures electrical activity associated with muscle contraction. In this system, surface electrodes detect activity from the Frontalis muscle above the eyebrow. Surface EMG is non-invasive: electrodes placed on the skin collect a combined signal from active motor units rather than recording from within the muscle.
Electrooculography (EOG) measures eye movement by sensing changes associated with the eye’s electrical dipole. Electrodes near the eyes detect signals as the eye moves horizontally or vertically. Both EMG and EOG are used as inputs for human-machine interfaces, including interfaces intended to support communication or control when conventional physical inputs are difficult to use.
These signals are small and susceptible to interference, so direct connection to a digital input is not enough. The acquisition chain must amplify and filter the electrode signals, manage common-mode voltage, and provide an output range suited to subsequent processing.
Wearable EMG and EOG device: acquisition-chain design
The reported system combines signal acquisition and conditioning in an analog circuit connected to a wearable mask. Its stages are designed to preserve the frequency range of interest while reducing unwanted low- and high-frequency components.
Amplification and common-mode interference
An INA128 instrumentation amplifier provides the initial amplification. Its high input impedance and common-mode rejection characteristics suit the task of measuring small differential signals from skin electrodes. Two 220-ohm resistors set the reported gain to approximately 114.63.
A right-leg driver circuit is also included. It feeds a common-mode signal back through a reference electrode to reduce common-mode interference at the body. This is an important design consideration in biopotential acquisition, where interference can be large in relation to the signal being measured.
Filtering and output conditioning
The circuit uses a high-pass filter with a cutoff between approximately 5 and 6 Hz to attenuate low-frequency components, including movement artifacts and other biopotentials. A second-order Butterworth low-pass filter, with a cutoff of approximately 120 Hz and unity gain, attenuates higher-frequency noise. Together, these stages form a bandpass response suited to the signal acquisition described in the study.
After filtering, a variable-resistance stage allows the gain to be adjusted, with a reported maximum of 13.19. The final conditioning stage adds a 2.5 V offset. This places the output within a 0–5 V range intended for use with an embedded system. Adjustable gain is useful because signal levels can vary; however, the gain must be set carefully to avoid saturating the output.
Electrode placement in the wearable mask
The mask holds five electrodes and helps maintain their positions during acquisition. Two electrodes above the right eyebrow record Frontalis muscle activity. Four electrodes are used for EOG: one above and one below the eye for vertical movement, and one at the outer area of each eye for horizontal movement. A connector lets the user select the electrode configuration for the desired signal.
In the reported tests, the EMG and EOG signals were acquired individually, even though the circuit was designed to support both. EMG recordings captured cycles of Frontalis contraction and relaxation. EOG recordings captured up-and-down and left-and-right eye movements, with voltage peaks appearing as the eye moved from its central position and returned.
The mask can connect to the circuit through an approximately one-meter cable. The authors noted that the cables were not shielded and that the system remained susceptible to electrical interference, including 60 Hz noise. Later acquisitions used LabVIEW with a DAQ USB-6212, with digital processing applied to address the interference. The wearable mask also helped limit unwanted electrode movement during recording.
What the circuit tests demonstrate
Bench testing used a 20 mV peak-to-peak input to evaluate the filter response and gain-adjustment stage. The measured filtering behavior was consistent with the intended cutoff range of approximately 5–120 Hz. Tests of the adjustable gain also showed why input amplitude and gain settings matter: increasing gain can drive the output toward saturation, particularly when test signals are larger than the biopotentials being measured.
These results demonstrate the operation of the acquisition and conditioning stages, not a complete industrial deployment or a finished control interface. The described EMG and EOG recordings were obtained separately, and digital processing was used to reduce 60 Hz interference. The work identifies simultaneous acquisition, a more compact circuit with additional channels, and embedded digital processing as areas for further development.
Implications for HMI and embedded-system design
For engineers evaluating biopotential inputs, the design highlights several practical points:
- Electrode geometry matters: placement determines which muscle activity or eye movement is captured.
- Conditioning must suit the signal: amplification, filtering, common-mode rejection, and offset all affect whether an analog signal can be used by downstream electronics.
- Gain requires calibration: adjustable amplification provides flexibility but must be balanced against output saturation.
- Interference needs a plan: analog circuit design, electrode and cable arrangement, and digital processing all influence signal quality.
These considerations apply to HMI prototyping and embedded acquisition work, but the reported system should not be treated as a ready-made industrial control solution. Its demonstrated scope is the acquisition of facial EMG and EOG signals under the described test conditions.
Conclusion
The wearable mask and its signal-conditioning circuit illustrate a complete path from skin electrodes to an embedded-system-compatible output: instrumentation amplification, common-mode management, bandpass filtering, adjustable gain, and a 2.5 V offset within a 0–5 V output range. For automation teams exploring hands-free or alternative HMI inputs, the main practical lesson is that reliable use depends on the whole acquisition chain—not just the sensor. Electrode placement, interference control, gain setting, and downstream processing must be considered together, while simultaneous operation and embedded processing remain areas for further development.