Measuring torso twist (scoliosis exercises) with IMUs. How to limit drift and test accuracy at home?

I'm a student building a wearable for people with scoliosis. Their home exercises need small, exact twists of the upper body relative to the hips, but nobody checks them at home. My idea: motion sensors measure the twist and give a sound cue like "more" or "less". Early stage, no prototype yet.
Rough plan (not final): 4 IMUs, 2 on the upper back and 2 on the hips.

Got some questions:

  1. How do I get the angle between two IMUs on one strap without drift building up over 10 minutes (filter, calibration pose)?
  2. 6-axis or 9-axis for slow movements indoors? Which IMU under ~$15 would you pick?
  3. How would you connect 4 IMUs to one ESP32 (I2C multiplexer, SPI)?
  4. How can I check angle accuracy at home without a lab?
  5. Is there open-source code or a library that computes the angle between two IMUs on an ESP32 that I can start from?

A one line answer or a link helps. I'm also looking for 1–2 people who'd like to review my ideas now and then, or join later as a collaborator (student project, unpaid). If that's you, message me, I'm open!
I'm Viktoriia, from Ukraine

Comments

  • For the accuracy check, I’d start with two sensors on the same rigid board. Turn the board slowly, then leave it still. The relative angle should be zero throughout. How far does it wander in ten minutes? Next, keep one sensor on the table and turn the other on a board with a pointer and a printed protractor. Approach each mark from both directions. Repeat after removing and remounting the sensors. That gives you separate estimates for drift and mounting repeatability before strap movement or the shape of someone’s back enters the measurement.

Sign In or Register to comment.