Bench modeSteps, parts, and safety only. Big type for a phone at the bench.
Microvolt
Phase 5: Build for a personProjectA semesterAbout $100 beyond the amplifierTier 2Needs a lab

Project A: A speller for someone with ALS

Hybrid EOG plus SSVEP, designed with a real partner found through the ALS Association or the rehabilitation hospital. Start with the eyes, add EEG for when the eyes fail. That is the clinical progression, and it is the project that writes your internship essays.

AssumesProject D: Eyes as a joystickProject A: The SSVEP spellerNeeds findingSpineClinical / regulatory / human factorsDecoding / signal processing / MLAnalog / mixed-signal hardware

You are skimming: the title, the first figure, and the short version. Switch to Read in the header for the full page, or Deep to open every deep dive.

takes movement and eventually speech while leaving thought intact. Communication devices for people with ALS start with switches, move to eye-gaze cameras, and, when the eyes fail too, have almost nothing left; that is where BCIs are aimed. In this project you find one person, through an organization, learn what they actually need, and build a speller that starts with the EOG joystick from Phase 1 and can fall back to the SSVEP decoder from Phase 3. The person, not the technology, defines success.

Predict before you look

Commercial eye-gaze communication systems cost several thousand dollars and are very good. Why would a person with ALS want anything else?

Eye-gaze systems fail in exactly the situations that matter: bright light, ptosis, lying in bed, and the late-stage slowing of eye movements. Insurance approval can take months that the person does not have. The need is not “a better camera”; it is communication that keeps working as the disease progresses and in the bedroom at 3 a.m. Ask your partner and you will hear a version of this before you propose anything.

Finding the person

Through the ALS Association’s local chapter, a multidisciplinary ALS clinic (the University of Utah has one), or the rehabilitation hospital’s speech-language pathologists. Never by contacting patients directly. Explain that you are a student, that you want to learn what is needed, and that you can offer time and engineering. The speech-language pathologist is your most important ally; they know every device and every failure.

Listen first

Two or three visits before you build anything. Watch how the person communicates now. Time it. Notice what breaks: a caregiver has to recalibrate the camera every hour; the letter board is useless in the dark; the person has one message they need to send fast (“I need suction”) and it takes a minute. Write with no solution in them. Bring them back and ask which one matters most. The needs-finding explainer is the method.

The device

Stage one: EOG. Two channels of electrooculography, an amplifier with a very low high-pass, a small board on the headboard, Bluetooth to a tablet. The tablet shows a grid; horizontal and vertical gaze moves the highlight; a long dwell or a deliberate double-blink selects. Drift is the enemy: baseline tracking that re-centres whenever the eyes have been still, and a recentre gesture. Include a one-gesture emergency message that always works.

Stage two: SSVEP fallback. When gaze control fails, the same electrodes plus two occipital ones give SSVEP: four to eight flickering targets, a two-level letter tree, three to four seconds per selection. It is slower than gaze and it works when gaze does not. Photosensitivity screening is mandatory and is on the intake form.

Stage three: word prediction. Both stages become several times faster with a predictive text model that offers whole words after two letters. Any open-source predictive keyboard library works. Speed is measured in words per minute with the person, not characters per minute on you.

What makes it clinical

The design controls page has you write requirements (“selection error rate below 5 percent in the person’s bedroom lighting”), the FMEA page has you list every way it fails and what happens to the person when it does, and the human factors page has you watch a caregiver set it up without help. Do all three. The device is not done when it works for you.

  1. Contact the organization. Meet the speech-language pathologist. Get onto your lab’s protocol.
  2. Three listening visits. Ten need statements. The person picks.
  3. Build stage one on yourself. Then with the person, in their bed, in their light, for twenty minutes.
  4. Iterate on drift handling and the selection gesture until the person can spell a sentence unaided.
  5. Add the SSVEP fallback and test it with the person (screened) for a short session.
  6. Add word prediction. Measure words per minute and error rate over three sessions.
  7. Write the requirements, the FMEA, and the caregiver-setup observation. Hand over a device, a one-page guide, and your phone number.
Recall
Why does the project start with EOG rather than EEG?
Eye movement is the most reliable, fastest channel a person with ALS retains until late in the disease, and it works in conditions where eye-gaze cameras fail; EEG (SSVEP) is the fallback for when the eyes fail.
Recall
What is the single most important design requirement a listening visit is likely to reveal?
A fast, always-working emergency message matters more than spelling speed; the person needs to say a few critical things in seconds.
Recall
Why approach through an organization rather than directly?
Ethics and trust: the organization and the speech-language pathologist protect the person, know the existing devices and failures, and make an IRB-covered collaboration possible.