![](bg5.png)
5.3 Universal User
The results of the 4 test trials with 2, 2, 4 and 7
thoughts shown in testing data and results section
above are compared in Table 1 above.
From the results shown in Table 1, the universal
profile shows higher accuracy than the individual
profile in the majority of the test trials. Since the
subjects in the universal profile tests are using
similar muscle movements to create a specific
brainwave activity, these signals can be considered
to be consistent for different users. This is due to the
brain function to plan and execute these muscle
movements being similar across people (Jasper,
1949). These findings indicate the possibility of
creating a BMI system that can be used by the user
without any training.
5.4 Limitations of the Existing
Software
By comparing the unfiltered input sequences and the
desired input sequences, we discovered that the
Emotiv software handles the pure EEG signals and
facial expressions differently. For example, the
following is a set of data of the unfiltered sequence
and the desired sequences for a test trial.
The sequences and thoughts A, B, L, and R, are
represented in Table 10.
Table 10: Sequences and Actions.
A B L R Desired Sequence
Left Smirk Right Smirk Left Right BLABRA
The System unfiltered input is:
BR...LAAAALAAA...
Examining the two sequences, the user attempts
to obtain a left but he or she obtains a sequence of
LAAAALAAA. Since the left smirk and pure EEG left
thoughts have a significant interference region for
this subject, the user’s thought for pure EEG left is
also matched as a left smirk. The user is attempting
to obtain an L (Pure EEG Left) indicated by the
desired sequence. However, the sequence that the
system reads is LAAALAAA, indicating that the
user had executed more left smirk signals than pure
EEG left signals.
To examine this problem further, we analyzed
the frequency of which the Emotiv software
responds to the given signals and discovered a
potential reason that explained the abundant
matching of facial expressions compared to pure
EEG. In the Emotiv software, detection of facial
expressions and pure EEG signals are concurrent.
However in facial expressions, if a user kept their
facial expression consistent for a period of time,
EmoKey would repeat the keystroke to the robot
control program within that period of time. In
comparison for pure EEG signals, the signaled
keystroke would only be sent at the instant the user
executes a specific thought from the neutral state. A
future recommendation to improve this issue is to
test the facial expression and pure EEG detection
under the same detection system.
5.5 Disturbance Caused by Body
Throughout the test, we discovered that significant
body movements, speaking, or being in an excitingly
emotional state can produce a noisy signal and cause
an unintended action to be performed.
Figure 11: Brain Activity of Random Movement (Left) vs.
No Movement (Right).
In this analysis, a software called Emotiv Brain
Activity Mapping is used to illustrate the brain
activity when executing random body movements as
figure 24. Since the desired system ideally should be
able to function while the user is walking,
emotionally distracted, or even running, the system
should be able to filter out the noisy signals
generated by these movements, or emotional states.
Three possible solutions can be used to help address
this issue.
A) Implementing a Stop Feature: A simple stop
feature would halt the system from executing any
received keystrokes. This feature would also allow
the user to stop or restart the system easily such that
the user will not execute unintended actions while
they are idled from the system. Currently a stop
feature is already implemented in the current state of
the project, but in the future this feature can be
implemented as a facial expression based interface.
B) Sudden Strong Signal Filter: Since subjects
may generate very strong and noisy brain activity
when they are moving, an additional filter is needed
to stop the system from executing actions when it
receives a sudden, strong, and noisy signal. As a
result, this would only allow the system to accept
thoughts while the user is in a calm emotional state.
StudyofanEEGbasedBrainMachineInterfaceSystemforControllingaRoboticArm
343