During the data collection, the raw signals oscillated frequently
due primarily to noise signaling from the neck muscle and motion artifact. At
times, there was a distinct peak in the data, but the signal was not long
enough to represent a clear view of a typical EMG signal. Additionally, the
neck muscle signal is somewhat difficult to read using the equipment provided.
Although the ideal EMG signal shows much oscillation as well as shown in the
diagram, one could apply a curve over these values, which shows that there is a
distinct pattern in an ideal EMG signal. The muscle signal would show somewhat
of a bell curve where it increases, reaches a peak, and then levels back down
after the maximum signal is reached.
Figure 1
represents an ideal raw EMG signal where there is clearly a peak between 1.5 to
1.7 units
2. Describe motion artifact during
surface EMG/EOG collection. Where does it come from? What is the frequency
range of it?
Motion artifact is a form of interference that can lead to false
signals during data collection. This effect is caused by movement of the
sensors themselves rather than contraction of the muscles. What happens is that
since the sensors are sensitive enough to capture very minute surface EMG
signals they are also capable of registering the small amounts of static
electricity caused by the electrode sliding along the skin. Although these signals
are often very small, they can still lead to false readings because they look
very similar to small surface EMG signals. When we first began experimenting
with the data collection portion of this experiment we saw small oscillations
on the plot and initially thought that these were the EMG signals we were
looking for. We soon noticed that instead it was just motion artifact from the
other sensors being rubbed up against the table. The frequency range of the
motion artifact is from approximately -3V to 3V.
3. Other than motion artifact, what are some other sources of noise which can obstruct the surface EMG/EOG from being ideal? What are the corresponding frequency ranges?
Other sources of noise include applying pressure to the electrodes,
this causes the signal to jump from anywhere from -10V to 10V. Also, the
opposite effect occurs. When the sensors begin to not stick as well there can
be added noise in the form of the energy created from the sensors slowly
peeling off of the skin, that energy is enough to cause the signals to jump
from -3V to 3V. This makes sense because it is very similar to motion artifact.
The most drastic form of noise that can be formed is when the metal part of the
alligator clip comes in contact with the skin, this causes a rapid change from
positive 10V to -10V. One of the final ways that a noticeable amount of noise
can be introduced into the data is if the ground electrode falls off. When this
happens, it can cause a few different problems, it can cause the line to be
either at +/- 10V or it can cause a rapid oscillation of the data where the
line starts at 10V and then will rapidly fall to 0V and then quickly rise to
10V again repeatedly.
4. Using the information
from the previous two questions, what is the range of frequencies that the
bandpass filter should be set to accept?
Based on the information presented in the
previous two questions the bandpass filter should be set to accept frequencies
within a range of 3V to 8V and -3V to -8V. This is the optimum range for the
bandpass filter to function because it is the same as the working range of the
muscle signals being collected. Having the filter set to this will allow for
small signals caused by motion artifact which is between +/- 3V. This filter
will also block out data caused by things like direct contact with the sensors
and from a sensor falling off. The reason that +/- 8V was chosen was because
the intentional EMG signals are strong enough to make it into this range but it
leaves somewhat of a cushion for the upper bound in case there is a mistake and
a false signal doesn’t quite make it to the +/- 10V extremes.
5. If you are using EMG data, then describe the following:
a. What is the energy signal and why it is easier to apply a threshold to it rather than the raw or filtered EMG signal?
The energy signal normalizes the raw data. Since the raw data
oscillates and is negative at times, the energy curve negates these negatives
by squaring the values. The energy curve had no oscillation and came out to be
square waves. These made it easy to take the area under the energy curve. Additionally,
the energy curve is a representation of what the muscle is actually doing since
it eliminates the noise that was created by motion artifact. Commands in MATLAB
were used to filter the data so that the noise was no longer taken into
consideration, and the filtered data was used to create the energy signal
curve. The filter used set n=4 for the butterworth filter. Essentially, since
energy is multiplied by the change in time, this is an integral of the muscle
signal. It is easier to create a threshold from the energy signal because it
does not have negative values, and the curve is much smoother and does not
oscillate.
b. Describe in full how you are
relating the EMG data to the intended actions of your project.
The EMG data is an essential component in the project because
these data will determine how the linear actuator will respond to the neck
muscle. When the EMG data is collected, the energy curve can be used to set a
threshold. Since the EMG signals of the muscle movement downward is different
from the EMG signal of the neck muscle moving upward, the threshold can be set
to distinguish these two actions. Once these two actions are distinguished, the
linear actuator can be programmed using MATLAB to respond by moving upward when
the threshold is reached for the upward movement of the neck and vise versa for
the downward movement.
No comments:
Post a Comment