top of page
OUR DATA
Signals and FFTs
The plot to the right shows the sensor data overlaid with the moments that are intended to be flexing. The button was pressed and held for the duration of each flex and said button is shown on the right vertical axis.
We found that our sensor sees a flex as having consistently far greater magnitude than any of the noise that may come from other motions of the body.
In this plot, we see a comparison of the noise present in the signal output from the sensor under different circumstances. In these graphs, there are no flexes. We should, then, see a uniform horizontal line - ideally, there would be no variation in the voltage if the subject is not flexing. As we see, however, there is some noise present. When there was a lot of motion from the subject (high motion plot and dancing plot), the values seemed to waver a bit, whereas when still (low motion plot) the value was nearly constant. Additionally, we found that outside light did not significantly impact the readings; when a light was shined on and off of the sensor (“Bright” plot), we still saw a nearly constant value.
These plots show the frequency analysis of different noise environments while collecting data. As we can see, higher motion in the background corresponds to greater high frequency noise. Additionally, when flexing is involved, there seems to be a trend of higher magnitude near the 0 to 1Hz frequency range
In this plot, we see the frequency content of the signal while a flex was being held. This was performed because if a particular frequency was exceptionally prevalent while the subject was flexing, looking for areas where that frequency was present could prove helpful in identifying when flexes began and ended. It does seem that there are peaks between 0.5 and 1 Hz and 2.5 and 3 Hz in both plots, but it is hard to say if this is in fact more than simply a coincidence. Beyond this, there do not appear to be any frequencies that are especially noteworthy, and as a result the conclusions we can draw are limited. It is worth noting, however, that our sampling rate was very small. When sampling at only 20 Hz, we could only pick up frequencies less than 10 Hz. When we perform this with a faster microprocessor and a more rapid sampling frequency, it is possible that the same analysis could yield useful information.
Here, we have the collected data from the EMG sensor. The first dataset is a pretty simple sample consisting of a few flexes and relaxes just to give us some idea of what we can expect to see. On the bottom, however, we have a slightly more complex dataset. This was collected over a much longer period of time, and features flexes of different magnitudes, different lengths and even has the user bending their arm between flexes at about the 350 second mark. As we can see, this data is far more difficult to read in realtime. Noise often reaches the same thresholds as flexes, which makes analysis difficult without complex algorithms. Despite this, there exist clear trends and features present only when flexing, and these features can almost certainly be determined using DSP techniques.
bottom of page