Took things into my own hands and implemented movement detection from scratch using raw accelerometer data. Algorithm is based on calculating absolute "jerk" (derivative of acceleration) and recording it if it goes over a dead-band/threshold. Tried it on myself last night and results are much better! Can actually see patterns of tossing/turning vs still sleep in the graphs.
Discussions
Become a Hackaday.io Member
Create an account to leave a comment. Already have an account? Log In.