Today my wife and I tested out the Gyro Belt. We recorded once specific dance move, what we call the Alex head roll. (We don’t know its real name, the dance teacher who taught us was named Alex) The flat part of the graph, you’ll notice, starts pretty flat and then begins to pick up some noise. After that the large pulses come in. Although were not able to take video, the move entailed a moment of standing still while I counted beats in preparation, two complete Salsa basics followed at last by the multi-turn head roll move.
As you can see the move registered quite nicely on the graph. But this is a simple move in isolation. Later we’ll test the reading inside of an entire dance routine. I would expect the signal to noise ratio to drop substantially, but we’ll see.
Once thing I do notice, that is curious, is that the resting level at the end of the chart is different than the resting level at the beginning of the chart. That may give some trouble later depending on how much that zero rate level drifts. We’ll see.
Royce, what did you use to create the graph, and how many data points were there?
I used LibreOffice’s Calc program. I don’t remember how many datapoints there were. Quite a few though. The logger is supposed to sample at 50Hz and the move lasted around 20 seconds. I get a text dump of the data points in hex that I capture with a terminal program that connects to the uLog device. Then I wash that through a quick program that I wrote that converts the hex into tab delimited decimal numbers suitable for pasting straight into a spreadsheet.
Thanks… I need to dig into gnuplot a bit more. No other application could easily handle the accelerometer data I was throwing at it. (Sampling more than once per minute over many hours produces a lot of data!)