HolfordM_3
Kate and I are working to develop a device capable of detecting hypoglycemic trembling. We have gone about this task by programming a Raspberry Pi to read in and analyze accelerometer data. It is our thinking that hypoglycemia causes such a distinct trembling that it will leave a unique signature in the data.
In reality, this task is a bit more tricky. It’s funny that as a human we can look at someone trembling and someone chopping vegetables, and immediately know the difference. However, all the Raspberry Pi receives is three components of an acceleration vector, with absolutely no context.
We have tried to develop an algorithm to distinguish this trembling using two main routes. Firstly, we are doing research and reading papers to see if there is any “theory” we can apply. For example, metabolic tremors occur at 8-12 hertz, so we can calculate the Fourier transform and look at the frequency domain. Secondly, we are gathering data from our patient and simulating data ourselves to see if there is anything that can be discerned by trial and error.
Ultimately, saying “yes, this is trembling,” is not the hard part. It is eliminating the false positives that occur when other regular motions are picked up as similar to trembling.
Luckily, after countless hours of frustration, I [sort of ] finished the CAD assignment, so I should be able to dedicate a good amount of time to this algorithm in the coming days.
-Mira
BME Design 