Re: Occupation sensor that works with people sitting still?

["Bill Kearney" <wkearney-99@hot-mail-com>]

> I think the solution is going to be very much like our own multiple senses
> guide us to make decisions.  A combination of PIRs, threshold detectors in
> doorways, sound detectors, vibration sensors and others, wired into a
fairly
> sophisticated decision-making computer could do a reasonable job of
> assessing "occupancy."

I agree.

> The remarkable
> software that's finding its way into consumer cameras to detect faces and
> correct red-eye means it's not unreasonable to think tomorrow's occupancy
> sensor would be smart enough to know when people's eyes are trained on a
TV.

And of course the paranoid among us would worry about the 'gubmint' using
that to invade our privacy.  What's next, your DVR with facial recognition
being used by the feds to track down warrants?  Heh, talk about "Minority
Report".

> I've wanted a simple sensor to fit under chair legs that detects and
reports
> weight changes.

My new Jeep does that to control the airbag in the passenger seating.  But
then it complains if you put a briefcase on the seat.

> Someone sits down?  Send a signal that indicates a high
> level of probability that the user is still in the room, even if he
doesn't
> move.  If the occupant doesn't move for three days, call the coroner!

Help, I've sat and can't get up!  But yeah, since the population's aging it
would certainly be handy to have sensors dealing with health/safety.

> I'll
> bet I could make something out of a old Hawkeye, some casters and some
sort
> of piezo-electric strain gauge.  Plenty of space under or behind most
chairs
> and couches to velcro a Hawkeye and lead some wires down to the chair leg.

And drive the wife batshit with the all the crap attached to the
furniture...  I know mine would throw me outta the house if I proposed such
a thing.

What's really bad about all this is batteries.  All this junk would end up
requiring an endless loop of replacing batteries all the time.

> Precisely.  There was a very interesting occupancy sensor described here a
> while back that used radio waves.  It set up a "known pattern" with two
> antennas at opposite sides of the room (IIRC) and when that RF field was
> disturbed by humans beings interposed between them.

Yeah, that was my point about having to recalibrate.  It's sort of like the
old joke about teasing the blind by moving the furniture.  The RF pattern
setups would require the ability to be retrained in an unobtrusive way.

If you came into the room, moved the ottoman and plopped down on the couch,
what should the pattern be?  How would it be smart enough to know the
difference between than and, say, you and another person came into the room,
you sat on the couch and the other on the floor.

This is also where occupancy sensors fail because of pets moving about.  And
while you might be able to get Fluffy to wear an RFID tag it's not likely
you'd get people to do it.

The tracking necessary depends on so many factors as to be too expensive to
actually implement.  Day/night/electric lighting variations, heat/cooling
fluctuations, auditory, visual, infrared, etc.  By the time you factored it
all in, the expense of it would be well beyond what the market would bear
price-wise.


> I've got a similar set and any number of conditions are evaluated in the
> decision to turn it off.  There's progress being made, it's just very slow
> and uncoordinated.

There's the rub.  What's most tragic about all of this is most developer
geeks are too young/workaholic to actually have a home, or one they visit on
a regular basis.   They don't "GO HOME" or live the same sort of like
"normal" people do.  Thus it's a bit hard to get them thinking about
automating things in a real world sort of way.  No insult intended, that's
just the way things have been.  The people that write the code don't
actually have normal lives.

> That's a good plan.  In the two places I use X-10 Hawkeyes it would be
nice
> if they could start just ramping the controlled lamp off slowly instead of
> plunging the room into sudden darkness.  Hey, I could even do that via
> programming.  Good idea, dude!

Yeah, when you let the automation interact in a "stupid way" you avoid
having the people get pissed off when it fails to be "telepathic" to their
needs.  When people grasp the systems are 'dumb' they don't get their hopes
up, or beyond reasonable expectations about how it works.  Methodical,
simple and predictable automatic systems are viewed for what they are, not
what the users 'wish' they'd do.

I've been muddling around with it for a while.  It boils down to steps like:

is someone there?
is a 'person-related activity' in progress?
is it during a given time or type of day?
can I ask for guidance without bugging the person?

As in, did it detect your entry into the room, is the TV on, is it playing a
two hour movie, is it past your predictable bedtime, is it a friday night or
is tomorrow a holiday?  Then the trick is in finding an unobtrusive way to
let the occupant respond to override pending changes.  My office lights
blinking required getting up to flip the switch; not an unreasonable
requirement.  But that probably wouldn't fly for the couch potato watching a
movie.  The Tivo pops-up a message about it needing to change the tuner if
it senses you've been using the remote.  If you're not there or don't
respond to the prompt it automatically proceeds.  Using the overlay with the
remote is reasonably unobtrusive as you're already likely to have the remote
handy.  But then there's no decent API to let other systems interact with it
and doing overlay onto stuff like HDTV is not inexpensive.

> It's why we have ears, eyes, noses, nerve endings, etc.  One sensor is too
> easy to fool.  I remember back in sensory psych class learning about
> microsaccades.  There are tiny movements the eye constantly makes because
an
> image falling on the retina fades quickly.  If you can truly immobilize
the
> eye and stop the any saccadic motion, the image you see turns very quickly
> to gray.    Without microsaccades, staring intently at an object would
cause
> the image to vanish after a few seconds.  Rods and cones respond only to
net
> changes in brightness.  PIRs behave in a similar fashion.  As long as
there
> is motion, there's detection.  Maybe oscillating the PIR sensor is the
> solution! (-:

Excellent point.  Of course then there's also the problem of killing a fly
with a nuclear bomb.  Is that sensor too smart or too expensive to justify
it's cost?  Using RF in the home is going to be a real problem.  If not just
from the uncontrolled interference perspective but the human health impact.
Too many folks whinge about the risks of "cancer" from their cell phones,
the same complaints would undoubtedly get raised about occupancy sensors.
But then again, folks would whine about anything new...

> FWIW, you were probably right about the old refrigerator.  It sure does
suck
> down the kilowatts.  I should have a 7 day figure shortly.  The day to day
> variations were just too great to extrapolate from.  As my wife pointed
out,
> I have to factor in the cost of all the frozen stuff we buy on sale that
we
> wouldn't have room for if we didn't have the second box.  These efficiency
> calculations become very complex once you really begin to consider all the
> parameters.

Damn that inescapble logic!  Yes, it's not a bargain if it costs you extra
to keep it frozen.  When you've got a big family of growing children there's
some justification to it.  But otherwise it's sometimes cheaper to NOT get
it in bulk.  Go figure, literally.

> I'm not too worried yet.  As Dan pointed out, I am recapturing a lot of
that
> energy as heat and with subfreezing temps, that's a deal I can live with.

Heh, that reminded me of when my furnace died.  I was up in the office
working close to a Sun server (an ancient 4/3xx series chassis).  The box
radiated such a tremendous amount of heat I didn't even notice the HVAC had
died until I stepped away from the machine.  Literally about four steps.

-Bill Kearney