I decided to maximise the prevalence
of the piezoelectric effect (the ability of
certain materials to generate an electric
charge in response to applied stress).
This would steer away from the need
to use invasive methods or UV/
infrared screening since constant
exposure to EMF radiation is
detrimental to the user. The most
sustainable aspect of utilising the
piezoelectric effect is probably the lack
of need for disposal batteries or
charging cords.
When the user coughs, drinks or eats/
swallows, the piezoelectric vibration
sensor turns the vibration from
movements around the neck into
corresponding waves. After the data
reaches the microprocessor, spectral
analysis is done to convert the timebased
signals to frequency-based
signals. This can help identify sections
of the wave that correspond to specific
actions. It also removes anonymities,
perhaps from increased pulse rate or
background movements. In my
submission, I argued that Fast R-CNN
(a form of object detection architecture)
was most suitable to classify the action
from a wave, but I have since realised
that CNN would be more efficient
since there is no need to identify the
region of where the wave/object is.
Besides the piezoelectric vibration
sensors, piezo pressure and heat
sensors help identify changes in
pressure from falling or increases in
temperature due to a fever.
For this submission, I was highly
commended and was given a few
feedback suggestions. The judge was
concerned with ‘striking the balance
between the tightness of the choker
(comfort) and the ‘noise’ in the signal. ‘
They also mentioned that ‘Machine
learning could help but would need a
large amount of clean data to be
collected’. The feedback was incredibly
useful and will definitely be taken on
board to improve my future ideas.
From designing the wearable to
researching technologies for it, I found
the process incredibly fulfilling and I
learned a lot about solving real world
problems, beyond the logic and
systems we are taught in classroom
settings.
This was a truly stimulating
challenge.