Abstract:
The aim of this project is to obtain information about a user's sleep cycle and any present problems therein by use of
a non-invasive tremor sensor. In short, I wish to take a signal produced by a tremor, process it to a form that I can
extract meaningful data from (frequency, excursion length, etc.), and assign a label to the tremor.
The label assigned to the tremor would ideally be the disorder that caused it. I am looking into characterizing
tremors based on the cepstrum of the time domain signal obtained from two accelerometers attached to the hands.
The final product is a glove with an Arduino chip, accelerometer, and Bluetooth transceiver sewn onto it that will
transmit the position at a rate of 30 Hz to a computer running Matlab that will parse and analyze the data.
Challenges include preventing aliasing, identifying short impulses, and identifying tremors themselves from background noise.
Report (pdf)