The Uncertainty Tradeoff
A simplistic
algorithm was something that we had to come to terms with as we moved
toward the demo phase. Reason: We had run into something similar to
the Heisenberg Uncertainty Principle (not a Breaking Bad reference).
Simply, the more we worked to design our algorithm to recognize a
single movement pattern, the less able it was to recognize a broad class
of movements. The opposite was also true. The more we worked to
recognize a broad range of movements, the less capable the algorithm was
of recognizing specific patterns. We had to come to a happy medium
with the algorithm, keeping it simple enough to recognize a fairly
general class of movements, but, not so simple that it accepted every
possible movement pattern out there.
Workable Design
We
can only provide some specifics as we have to maintain some of our
trade secrets. By recognizing that deliberate movements are smooth and a
little slower, we were able to stabilize the accelerometer data against
all of the noise you all saw in our previous post. The next piece was a
method of detecting what position the phone came from and where we
wanted it to be at the end. All of these three pieces come together in
our first video demo
on YouTube. Realistically, we had one major goal at this time. We
wanted to see if different people could figure out how to lock and
unlock the phone with NO prior instruction. So, we rounded up some of our colleagues from OMNI (Ohio Musculoskeletal and Neurological Institute), coaxed them into giving our technology a go and be in the video. For this initial proof-of-concept demo, we created a "neutral" blue screen, a green "unlocked" screen, and a black "locked" screen. This initial demo, while not exactly pretty, shows that anyone can just pick up the phone and use it. Note: The demo smartphone is an HTC One, all of our OMNI folk are all iPhone users.
No comments:
Post a Comment