← All projects Project 001 / Assistive tech

LiDAR Glasses

Glasses with three depth sensors that warn blind and low-vision wearers about obstacles a cane can't reach, like branches and signs at head height, through vibration motors worn on the chest.

In progress
Sensors
3 time-of-flight
Range
Up to 5 m
Feedback
15 vibration motors
Compute
Raspberry Pi 4
Runtime
~8 h (estimated)
In editing

Build video coming soon

Subscribe to see the full build as soon as it's out.

Subscribe on YouTube →
The problem

A cane covers the ground, not the space above it.

A white cane is the best tool there is for the first metre or so of ground: curbs, steps, things on the floor. It can't reach branches, open cabinet doors, overhanging signs or the back of a parked truck, and it gives no warning about what's coming two to four metres ahead.

These glasses are built to work alongside a cane, not replace it.

How it works
  1. Sense

    Two wide depth sensors on the frame scan ahead, 120° across and up to 5 m out. A third, angled 40° down, watches the ground for curbs and drop-offs. A motion sensor tracks head tilt, so looking down at the floor doesn't read as an obstacle.

  2. Decide

    A Raspberry Pi 4 in a pocket merges all three sensors into one depth map. When two sensors disagree, it trusts the nearer reading, because warning too early is the safe mistake.

  3. Feel

    A 3 × 5 grid of vibration motors on the chest. Columns show direction and rows show height: trip level, torso, overhead. The closer something is, the faster it pulses.

Design choices
Fewer signals
Only the one or two most urgent obstacles drive the motors. A full depth map on 15 motors becomes a constant buzz that the brain tunes out within a minute.
Pulse speed
Urgency is shown by how fast a motor pulses, not how hard. People can only tell a handful of vibration strengths apart, but a quickening pulse is hard to ignore.
Real height
The rows follow an obstacle's actual height, not where your head is pointing, so a warning doesn't slide up and down your chest when you nod.
Fail silent
If the software ever stalls, the motors switch off within half a second instead of holding an old warning.
Progress
Done
  • Sensor fusion software, tested in a full simulator
  • Head-motion tracking
  • Drivers for all three sensors
  • Vibration pulse system
  • First depth sensor running live on the Raspberry Pi
Now
  • Bringing up the second depth sensor, the ground sensor and the motion sensor
Next
  • Wiring the 15-motor vibration array
  • Choosing the final sensor layout and 3D printing the frame
  • First walking tests
Get involved

If you're blind or have low vision, or work with people who are, your input on what would actually help matters more than anything else this project can get.

connect@debacktech.com