Group Actions
Engagement
  • Author Author: khadri
  • Date Created: Date Created
  • Views 47 views
  • Likes 2 likes
  • Comments 2 comments
Related
Recommended

EyeLink — A Wheelchair You Drive by Blinking

khadri
thumbnail

Head Tilt + Eye Blink + Wireless Control

I built EyeLink, a working prototype of a hands-free wheelchair control system that combines head movement and eye-blink confirmation to control a small motorized wheelchair platform.

Instead of using a traditional joystick, the user tilts their head to select a direction and then uses a specific blink pattern to confirm the command.

The important part is that neither input is trusted by itself.

A head movement alone cannot start the wheelchair, and a blink alone cannot start it. The direction must first be selected and then confirmed before a movement command is transmitted.

The wearable controller uses an ESP8266, MPU-6050 and two IR blink sensors. A second ESP8266 mounted on the wheelchair receives the confirmed command through ESP-NOW and controls the motors.

This project is a working prototype and proof of concept, not a certified medical wheelchair.

final test

1. The Idea

The idea came from combining two different types of sensing:

  • Eye-blink detection
  • Head-tilt detection

I wanted to see whether these two signals could work together as a deliberate control interface.

The basic sequence is:

Head tilt → Direction selected → Blink confirmation → Wireless command → Wheelchair movement

This two-stage approach became the main design principle of EyeLink.

A random blink while the head is centered should do nothing.

An accidental head movement without the required blink confirmation should also do nothing.

test1

2. Hardware

Wearable Transmitter

The glasses contain:

  • WeMos D1 Mini / ESP8266
  • 2 × IR reflective blink sensors
  • MPU-6050
  • Push button
  • LiPo battery
  • TP4056 charging module
  • 3.7 V → 5 V boost converter
  • Safety-glasses frame

The MPU-6050 detects the orientation of the user's head, while the two IR sensors detect deliberate blink patterns.

I kept the electronics compact because the transmitter has to be worn on the user's head.

Wheelchair Receiver

The receiver contains:

  • WeMos D1 Mini / ESP8266
  • L298N motor driver
  • DC geared motors
  • HC-SR04 ultrasonic sensor
  • 2 × 18650 cells
  • Power LED
  • Status LED
  • Power switch
  • Activation button
  • PVC foam-board chassis

3. Building the Wearable Controller

I started with a lightweight pair of safety glasses and mounted two IR blink sensors inside the frame.

The sensor position is important because the distance and angle relative to the eyelids affect the detection response.

During testing, I also found that strong ambient infrared, especially direct sunlight, can affect the sensors. This is one of the limitations I plan to address in a future version.

I then added the ESP8266 and MPU-6050.

The MPU-6050 is mounted on the glasses and is used to detect:

  • Forward
  • Backward
  • Left
  • Right

eyeblink sensor

esp tx

mpu

4. Improving the Power System

During early testing, the blink sensors and MPU-6050 were not responding as reliably as expected.

After investigating the problem, I found that the available battery voltage was not suitable for stable sensor operation.

I added a small boost converter to provide approximately 5 V to the sensor/control section.

After this change, the sensor response became much more stable.

This was one of the practical lessons from building the prototype:

theoretical component specifications are not enough — the complete power system has to work reliably together.

battery

5. Building the Wheelchair Platform

I built the drive base using PVC foam board and installed four geared DC motors.

Initially I tested 300 RPM motors, but they were too fast for the small prototype.

The movement was abrupt and turning/stopping was difficult to control.

I changed to 100 RPM geared motors, which produced much more controllable movement.

This was an important design decision made from actual testing rather than simply choosing motors from their specifications.chasis

motors and driver

6. Obstacle Detection

The user's head and eye signals tell the system what movement is intended, but they do not tell the wheelchair whether an obstacle is physically in front of it.

So I added an HC-SR04 ultrasonic sensor to the front.

In the current implementation, forward movement is blocked when an obstacle is detected within approximately 20 cm.

The current implementation uses a hard stop rather than gradual speed reduction, so this is an area I want to improve in the next version.

ultrasonic

7. Wireless Communication — The “Message” in Make A Connection

This is where EyeLink directly connects with the Make A Connection theme.

The glasses and wheelchair communicate using ESP-NOW.

The communication path is:

Eye/Head Sensors → ESP8266 TX → ESP-NOW → ESP8266 RX → Motor Driver → Motors

ESP-NOW allows the two ESP8266 devices to communicate directly without:

  • Wi-Fi router
  • Internet connection
  • Smartphone application
  • Cloud service

The wearable unit creates a command, and that command is transmitted as a wireless signal to the wheelchair receiver.

So the project is not simply detecting a gesture — it is detecting, confirming and transmitting a control message wirelessly

8. Software Setup

logic

The software setup follows this sequence:

Get RX MAC Address → Calibrate MPU-6050 → Configure TX → Upload RX Firmware

First, I obtain the receiver ESP8266's MAC address.

Then I calibrate the MPU-6050 for the actual physical position of the sensor on the glasses.

The calibration records the normal position and the four directional positions:

  • Forward
  • Backward
  • Left
  • Right

The final system uses two main firmware programs:

TX_Glasses_FINAL.ino

and

RX_Wheelchair_FINAL.ino

The complete firmware and supporting files are available with the project documentation.

https://github.com/humanixtechlab/EyeLink-Wheelchair.git

9. The Core Control Logic

This is the most important part of EyeLink.

I did not want one sensor event to directly control the motors.

Instead, the firmware works in two stages.

Stage 1 — Select

The user tilts their head in a particular direction.

The system checks that the movement remains in the required position and locks the selected direction.

Stage 2 — Confirm

Only after the direction has been locked does the system look for the correct blink pattern.

The logic is:

Head Tilt → Direction Lock → Blink Confirmation → Movement Command

Both conditions must agree before the command is accepted.

This is the central control concept behind EyeLink.

10. Movement Commands

Forward

Tilt head forward + double blink with both eyes

forword

Backward

Tilt head backward + double blink with both eyes

Forward and backward can continue moving after the command has been confirmed, making longer movements more practical.

backword

Left

Tilt left + double blink with the left eye

Right

Tilt right + double blink with the right eye

Left and right are handled as shorter movement pulses rather than continuous rotation, giving better control during directional changes.

left

11. Communication-Loss Protection

I also wanted the system to stop if communication between the glasses and wheelchair was lost.

The receiver has a communication timeout.

If expected communication from the transmitter stops, the receiver enters the stopped state rather than continuing to drive the motors.

During testing, I discovered an unexpected problem.

When the system was active but the user wasn't giving a movement command, the transmitter wasn't sending packets. The receiver interpreted this silence as a communication failure.

I solved this by adding an active-idle heartbeat.

While active, the transmitter periodically sends a small status packet even when there is no movement command.

The current heartbeat interval is approximately 300 ms.

power on

hold

12. Final Prototype

After the electronics and control system were working, I finished the physical appearance of the prototype.

I used:

  • PVC foam board
  • Sponge/foam
  • Bike-seat-cover material
  • Spray paint

This doesn't change the electronics, but it makes the prototype much easier to understand visually and gives the final build a more finished appearance

open case.

seat cover

13. Testing the System

I deliberately tested incorrect inputs as well as normal commands.

Examples included:

  • Blinking while the head was centered
  • Using the wrong blink pattern
  • A single blink instead of the required pattern
  • Moving the head without confirmation
  • Communication loss
  • Obstacle detection
  • Forward movement
  • Backward movement
  • Left movement
  • Right movement

The expected result for an incorrect gesture is that the wheelchair remains stationary.

The basic successful sequence is:

Tilt → Hold → Blink → Command Confirmed → Wheelchair Moves

I then tested the complete system with the glasses, wireless communication, motor control and obstacle detection working together.

left

test1

14. What Makes EyeLink Different?

The main engineering idea isn't simply “control a wheelchair with blinking.”

The interesting part is how the signals are combined.

Two-factor movement confirmation

Head movement selects the intention.

Eye blinking confirms the intention.

Neither input is sufficient by itself.

Another interesting design decision was repurposing inexpensive IR blink-sensing technology for a completely different control application.

I also chose ESP-NOW instead of a router/cloud/app-based system, keeping the communication local and direct.

These design choices came from the actual development and testing process rather than being added only for presentation.

final 2

final3

15. Limitations

EyeLink is a working prototype, not a certified medical mobility device.

There are several limitations that I want to be clear about.

IR sensor sensitivity

Strong ambient infrared, especially direct sunlight, can affect blink detection.

Battery protection

The current receiver uses a 2S Li-ion battery pack, but this prototype does not currently have a BMS. A proper 2S BMS is planned for the next revision.

Chassis

The current PVC chassis is a prototype drive platform and is not intended to be the final structural frame of an everyday wheelchair.

Obstacle handling

The current obstacle system uses a hard stop. Smoother speed reduction and more advanced obstacle handling are future improvements.

Wearable design

The glasses enclosure and sensor mounting can still be improved for comfort, reliability and repeatability.

These limitations are important because I want to present EyeLink as what it actually is: a working engineering prototype that can be developed further, rather than claiming that it is already a finished medical product.

16. What I Learned

The project went through several practical changes during development.

I:

  • Changed the motors from 300 RPM to 100 RPM
  • Added a boost converter after discovering unstable sensor behaviour
  • Calibrated the MPU-6050 for the actual glasses
  • Developed the two-stage gesture logic
  • Added communication-loss protection
  • Discovered and fixed the active-idle heartbeat problem
  • Added obstacle detection
  • Tested incorrect gestures deliberately

For me, these changes were as important as the final working prototype because they showed where the real engineering problems appeared after the first version was built.

17. Future Improvements

The next version could include:

  • Better glasses enclosure
  • Improved IR sensor shielding
  • More robust sensor mounting
  • Better gesture calibration
  • Proper 2S BMS
  • Stronger wheelchair frame
  • Smoother obstacle handling
  • Further communication and safety testing
  • Improved wearable comfort

EyeLink is therefore not the final version of the idea. It is a working prototype and a foundation for further development.

18. Conclusion

EyeLink combines head movement, eye-blink confirmation, and wireless communication into a hands-free wheelchair control concept.

The most important idea is simple:

One signal selects the intention.
Another signal confirms it.
Only then is the command transmitted.

The project demonstrates how relatively low-cost embedded hardware can be combined into a human-machine interface, while a direct ESP-NOW wireless link connects the wearable controller to the wheelchair drive system.

The complete project is open source, including the TX and RX firmware, MPU-6050 calibration firmware, receiver MAC-address finder, wiring/pin information, build documentation, and project images.

Link Project Resources

rx

tx

Busts in silhouette Humanix Tech Lab Team

Humanix Tech Lab is a maker and engineering team focused on developing practical DIY electronics, embedded systems, robotics, and assistive-technology projects.

Team Members

  • Khadri — Co-Founder & Engineering / Hardware
  • Nookaraju — Co-Founder & Engineering / Software & Documentation

Humanix Tech Lab — DIY Electronics • Embedded Systems • Robotics • Innovation

The project was built and documented by Humanix Tech Lab.

I built EyeLink as a low-cost proof-of-concept, but I see the larger idea as a reusable control interface: one input to select an intention and another input to confirm it.

There is still plenty of room to improve the hardware, safety systems, wearable design, and real-world reliability. This prototype is intended as a foundation for further development rather than a certified medical device.

Thanks for reading!

— Humanix Tech Lab

Parents Comment Children
No Data