9 Computational EEG Analysis for Brain-Computer Interfaces
195
the task. Selection of one or more of these options in sequence allows the user to type
messages, navigate menus, engage preprogrammed actions of an external device, etc.
Continuous mouse-like control is achieved by translating the coordinated modulation of the user’s brain activity into the intended directional control commands. This
allows the user to navigate a cursor to icons on a computer screen, freehand draw, and
can be directly extended to achieve continuous control of a robotic arm, a wheelchair,
or other devices that require continuous dimensional control.
Although achieving reliable 1-, 2-, or 3- degree of freedom control is useful and
sufficient for most assistive applications, a more ambitious objective is to design a
BCI that attempts to replicate natural, high degree-of-freedom function in a manner
that is more transparent to the user. For example, for communication, language cortex
signals during imagined speech would be decoded by the BCI and replicated using
a speech synthesizer in real-time. Similarly, for motor control, motor cortex signals
during imagined limb movements would be decoded by the BCI and replicated using
a prosthetic limb, orthosis, or even neuromuscular stimulation of the impaired limb
in real-time.
Approaching transparent replication of natural function via a BCI proves to be
very challenging for a variety of reasons. For instance, given the distributed complexities of motor and language processes in the brain, it is difficult to capture all
of the subtle nuances needed to reliably reproduce completely natural function from
limited recording sites (even on the order of thousands of single neuron or local
field potential recordings). Related to this point, it is expected that only invasive
recordings can provide the appropriate signals and resolution required to achieve
this type of high-level intrinsic control. Additionally, other difficult issues such as
the role of sensory feedback and proprioception in the replication of natural function must be considered. Consequently, simplified approaches that require fewer
recording sites and less sophistication, such as limited vocabulary speech and limited degree-of-freedom motor commands, can serve as more practical alternatives as
the technology continues to advance.
Any of a variety of brain signals can be translated by a BCI to achieve a particular
device output. For example, brain activity from relevant language areas recorded
while a user imagines vocalizing a word can be used to control a speech synthesizer.
This same brain activity could also be used to control a hand orthosis, where the
presence of a particular imagined word would close the orthosis and another imagined word would open it, for instance. Another example is a BCI that records from a
single neuron (not necessarily from motor cortex) that has been conditioned to adjust
the spike firing rate when the user wants to open/close the orthosis. Likewise, a signal generated over the sensorimotor cortex during imagined hand movement could
be decoded by the BCI and used to open/close the hand orthosis corresponding to
the imagery. This same sensorimotor signal could be used by a BCI in a communication application to select sequentially highlighted letters in a visual keyboard at
the moment when the imagined movement is detected. The resulting typed message
could also be synthesized as speech, thus achieving the same effective output as the
first example using a completely unrelated signal and interface.
195
the task. Selection of one or more of these options in sequence allows the user to type
messages, navigate menus, engage preprogrammed actions of an external device, etc.
Continuous mouse-like control is achieved by translating the coordinated modulation of the user’s brain activity into the intended directional control commands. This
allows the user to navigate a cursor to icons on a computer screen, freehand draw, and
can be directly extended to achieve continuous control of a robotic arm, a wheelchair,
or other devices that require continuous dimensional control.
Although achieving reliable 1-, 2-, or 3- degree of freedom control is useful and
sufficient for most assistive applications, a more ambitious objective is to design a
BCI that attempts to replicate natural, high degree-of-freedom function in a manner
that is more transparent to the user. For example, for communication, language cortex
signals during imagined speech would be decoded by the BCI and replicated using
a speech synthesizer in real-time. Similarly, for motor control, motor cortex signals
during imagined limb movements would be decoded by the BCI and replicated using
a prosthetic limb, orthosis, or even neuromuscular stimulation of the impaired limb
in real-time.
Approaching transparent replication of natural function via a BCI proves to be
very challenging for a variety of reasons. For instance, given the distributed complexities of motor and language processes in the brain, it is difficult to capture all
of the subtle nuances needed to reliably reproduce completely natural function from
limited recording sites (even on the order of thousands of single neuron or local
field potential recordings). Related to this point, it is expected that only invasive
recordings can provide the appropriate signals and resolution required to achieve
this type of high-level intrinsic control. Additionally, other difficult issues such as
the role of sensory feedback and proprioception in the replication of natural function must be considered. Consequently, simplified approaches that require fewer
recording sites and less sophistication, such as limited vocabulary speech and limited degree-of-freedom motor commands, can serve as more practical alternatives as
the technology continues to advance.
Any of a variety of brain signals can be translated by a BCI to achieve a particular
device output. For example, brain activity from relevant language areas recorded
while a user imagines vocalizing a word can be used to control a speech synthesizer.
This same brain activity could also be used to control a hand orthosis, where the
presence of a particular imagined word would close the orthosis and another imagined word would open it, for instance. Another example is a BCI that records from a
single neuron (not necessarily from motor cortex) that has been conditioned to adjust
the spike firing rate when the user wants to open/close the orthosis. Likewise, a signal generated over the sensorimotor cortex during imagined hand movement could
be decoded by the BCI and used to open/close the hand orthosis corresponding to
the imagery. This same sensorimotor signal could be used by a BCI in a communication application to select sequentially highlighted letters in a visual keyboard at
the moment when the imagined movement is detected. The resulting typed message
could also be synthesized as speech, thus achieving the same effective output as the
first example using a completely unrelated signal and interface.