Today is a big day for us. After two Small Business Research Awards, one
from DoD and one from USDA (which ended this August), we are launching the
Beta testing Phase of our Honey Bee Acoustic Scanner, starting with a
presentation this morning at the Montana Beekeepers Association Meeting in
Bozeman by Robert Seccomb, the Principal Director for this project, and one of
my research and business partners.
I really wanted to be there, but some bug hit the family this week, and
I've had to stay home. Call me Nurse Crabby.
We've struggled with making this device affordable - and we finally made
the breakthrough when a new technology became available in July. The good
news is that we managed to use this to put together a prototype device.
Bad news, we're out of money, just as it is time to calibrate the unit. As
of current date, its about 86% accurate for several variables like varroa
mites, Nosema, queenless. Does really well for Africanized bees - but
that's based on a really small set of AHB recordings.
We accomplished two things this summer - produced a working prototype, and
expanded the acoustic sampling system from the audible (to humans) into the
ultrasonic (up to 75,000 HZ).
Moths jam bat sonar by producing ultrasonic sounds, and we expect to find
some mechanical sounds in this range from activities such as chewing by
beetles. We've already found some ultrasonic sounds (possibly harmonics) from
queens. We also suspect that there may be less background environmental
noise in the ultrasonic range deep inside a beehive.
In practice, one walks up to an undisturbed hive, slides in a 12" long,
1/8" diameter probe microphone and take a recording. For the Beta phase
testing, we're asking for a two minute recording. In use, when trained, the
device only requires a 30 second recording.
The system identifies pests and diseases, based on the sounds produced by
the colony, using Artificial Neural Networks (ANNs). That's a fancy way of
saying the computer programs use pattern recognition - same system used to
process human speech. The more recordings of known variables (such as
presence of mites at specific levels), the better it gets. In fact, not only
do we train on multiple recordings, we also use multiple training systems
(networks of ANNs all processing the data together).
For the Beta testing, we are conducting a pre-sale of the device. We will
sell it for our cost (~ $650) plus shipping, and will supply two 32 GB SD
cards. We want experienced beekeepers and researchers. We ask each
participant to record colonies, then inspect and quantify where possible
conditions like mite presence, levels. We're working on a checklist. Then, when
a card is full, send it with the colony report to us. Robert Seccomb, who
is my research partner and developer of the unit, will use these data sets
to retrain and upgrade the system. The participant will get an empty SD
card back, and periodic updates to the software.
As an example, the device is currently scoring about 96% on Africanized
Bees, but that's based on a small number of very African hives. If we can
get recordings of 100 or even 1000 AHB colonies, the pattern recognizing
software can only get better.
So we plan on conducting a year-long Beta testing by real customers in
their real environments prior to the release and sale of the device. The very
good news is that we had 7 years of funding from DoD and USDA to work out
the kinks, produce the prototypes. We've a draw full of units that are now
obsolete. We've gone down many exploratory paths that resulted in no
improvement or even failure.
The prototypes are being hand built, so the cost is still higher than I
like. If, in the Beta testing, we could get orders for 100 units, the
supplier of some of the main components will cut the cost of the parts by as much
as 26%.
Since we're out of money, so to speak, I'm looking for funding to cover the
cost of cleaning up (for the training sets, we have to remove background
sounds) the data sets, entering them into our library, then processing with
the ANNS, and finally upgrading the algorithms in the device). Again,
looking to cover costs, so we're going to ask - if we sell this to you at
cost, would you donate something for the data processing and device updates?
Maybe an annual subscription?
Bottom line, we could go back through the competitive grant process, try to
find funding for the Beta stage, but that will delay everything by another
year. Or, we can go to the potential users.
The more units, the more verified data files, the faster they come in, the
quicker this system will convert from a research prototype into a useful
tool for bee management.
You can see what the prototype looks like on my personal Facebook page
(jerry bromenshenk). Nice thing about having a unique name, there's only
one on Facebook. We're working on a web page for the Beta testing.
Thanks
Jerry
***********************************************
The BEE-L mailing list is powered by L-Soft's renowned
LISTSERV(R) list management software. For more information, go to:
http://www.lsoft.com/LISTSERV-powered.html
Guidelines for posting to BEE-L can be found at:
http://honeybeeworld.com/bee-l/guidelines.htm
|