Wearable Fall Detection for Independent Living
A wrist worn device that distinguishes a genuine fall from sitting down heavily or dropping the device, and raises an alert with a cancellation window. False alarms are what get these devices taken off and left in a drawer, so that is the metric to optimise.
Advanced
Needs a team, real planning, and a supervisor who knows the area.1 semester, 2 to 3 students
5 to ship
3 optional extrasSuggested stack
What you should ship
- Wearable capturing accelerometer and gyroscope data at a rate sufficient for impact detection
- Labelled dataset covering simulated falls and a wide range of daily activities that resemble them
- On device classifier running within the memory and power budget, with measured battery life
- Alert flow with an on device cancellation window before any contact is notified
- Evaluation reporting sensitivity and, prominently, false alarms per day of normal wear
If you have time left
- Inactivity detection as a secondary signal after a suspected fall
- Gait change monitoring over weeks as a fall risk indicator
- Fall location estimation to include in the alert
The problem
A fall that goes undiscovered for hours is a major cause of poor outcomes for older people living alone. Detection devices exist and are frequently abandoned, almost always because of false alarms.
What you build
A wearable, a labelled dataset, an on device classifier, and an alert flow with a cancellation window.
The metric that determines success
False alarms per day, not sensitivity. A device that catches every fall and fires twice a day during normal activity will be removed within a week, at which point its sensitivity is zero. Report false alarm rate first and treat sensitivity as the constraint.
The data problem, and how to handle it ethically
You cannot ask older people to fall over. Standard practice is simulated falls onto crash mats by younger volunteers, with the limitation stated honestly, since real falls in older adults have different dynamics. The activities of daily living data matters more anyway, because that is where false alarms come from, and you can collect that realistically.
Why on device inference is the right choice
Streaming continuous motion data over Bluetooth destroys battery life and creates a privacy problem. A small quantised model on the device sends only events, which is better on both counts.
The design feature that matters
A cancellation window. Most alerts are false, and letting the wearer cancel before anyone is contacted is what makes an imperfect detector acceptable to live with.
Scope warning
Detection and alerting. Not a care platform, not medical records.
Ideas and guidance, not finished projects
These are project ideas and scoping guidance, published free for students to use as a starting point. I do not build, write, or sell final-year projects, and I do not complete coursework for anyone. Take an idea, make it yours, and build it.
Not sure this one fits you?
I help students pick an idea that matches their skills and their deadline, then scope it down until it is finishable. Guidance only, never done for you.
Stuck on something specific?
Writing only gets you so far. If you want an answer to your situation rather than the general case, book a session and we will work through it together. Every session is free; a few slots open each week.
Follow along
New writing, resources and project ideas land here first.