Biomedical Engineering Reference
In-Depth Information
the application, as dictated by your requirements. Design is about tradeoffs. Know
which ones can be made.
16.4.4 Find How People Want to Interact
A developer may follow their intuition to rapidly iterate on prototype systems, but
this method is costly and leads to ambiguous results:
Did the prototype fail because of
the design or because the recognition was poorly implemented
? A method to avoid
this ambiguity while exploring the design space is the Wizard of Oz protocol [
3
]
where a human operator interprets a user's actions (if an operator observing the user
cannot recognize their intent, a computer probably can't either). In this way, several
“prototypes” can be tested without time spent implementing recognizers. As well,
cleaner results can be obtained as the operator can adapt to a user as they explore
how they would move for all the actions needed in an environment and as a design
solidifies, the protocol allows for capturing sample data. This method allows users
complete freedom to: choose their movements for their intent, create new actions
on-the-fly and adapt as new requirements of the system are given that would conflict.
The operator also sees how fatigue affects the user. Video recording analysis, post-
experiment interviews or even data returned from devices you gave the user during
the Wizard of Oz session can inform the development of the recognition algorithms.
This approach was used on a commercial video game Mirror's Edge [
13
].
Mirror's Edge is a first person action-adventure video game incorporating Parkour-
like methods of travel as well as basic fighting motions. Mirror's Edge requires the
player to run, jump, climb and duck as they travel between buildings and across
rooftops in an urban landscape. Using the Wizard of Oz protocol, the operator
watched a real-time video feed, interpreted the player's actions and controlled the
game accordingly with a keyboard and mouse. The result was seamless: not all play-
ers even realized there was a human in the loop. From this, a generalized gesture set
for locomotion was created [
13
]. These same actions are common to many locomo-
tion systems and can form the starting point for most systems, having already been
incorporated into trainers [
27
,
28
] (Table
16.1
).
16.4.5 Compensate For Technology Limitations
Once you have your devices, recognizers and interface design, which works to min-
imize errors, you still can expect recognition errors and the impact of this can be
minimized through design. One approach is to focus improvements on the most
commonly used and most critical actions. For instance, a military trainer should
have walking be responsive, shooting triggered by a button and the grenade throwing
recognizer minimize false positives (i.e. not randomly toss a grenade at your feet).
The following three approaches offer non-technical compensation strategies.
