0:09
- You know how when you walk outside with barbecue meat,
0:12
the flies manage to find you almost immediately?
0:15
While this may be annoying for us,
0:16
it's very useful for forensic experts.
0:18
Flies can smell the meat
0:20
and therefore locate a dead body within minutes.
0:22
And in doing so,
0:23
they act as a stopwatch for the time of death.
0:26
Certain species of flies are better than others
0:29
at locating the smell and quickly laying
0:30
the infamous maggots that feed on the flesh.
0:33
So what we often see is a sequence of species
0:35
arriving one after the other in a somewhat orderly fashion.
0:39
Depending on the species of flies and maggots
0:41
present at a time, we can backwards calculate
0:44
to the day they first arrived,
0:45
and therefore, the day the person has likely died.
0:48
But some species of flies are more effected
0:51
by the environmental conditions.
0:52
So depending on the season and location,
0:55
the order of fly arrival may vary.
0:57
As such, time of death estimations using fly arrival data
1:01
still remain very approximate
1:02
despite being wildly researched.
1:05
But the current method of studying
1:06
the order of fly arrival has a few challenges.
1:09
It is invasive and it is irregular.
1:11
So depending on the interval
1:13
and the amount of data collection,
1:15
it can significantly affect the natural processes.
1:18
It is also time-consuming,
1:19
which leads into the lack of practitioners,
1:22
and it's not fully applicable to bodies that are buried,
1:25
covered, or enclosed,
1:26
such as those we normally found in car trunks.
1:30
This is where we step in.
1:31
We are currently testing an alternative method
1:33
to study the order of fly arrival
1:35
using the buzzing sound of flies.
1:37
There is enough evidence to suggest the flies
1:39
produce slightly different buzzes
1:41
determined by the combination of body and wing size.
1:43
In fact, us humans can hear their differences
1:46
if we really focus and if we have the sound file
1:49
right next to each other,
1:50
but machine learning algorithms can do it
1:52
much better than us,
1:53
and they have been used to discriminate
1:55
between mosquito species for malaria prevention
1:58
and other insects for border control.
2:00
We're currently creating a database
2:02
of forensically important flies
2:03
recorded under different conditions
2:05
in order to train a machine learning algorithm.
2:07
The conditions that we are focusing on at the moment
2:09
are temperature, humidity, age, size, and sex.
2:12
We're also recording the sound of flies
2:14
present on piglet experiments and human donors
2:17
at the Australian body farm,
2:18
which is a facility for decomposition studies.
2:21
We want to use these field recordings to test our algorithm
2:24
and check whether it'll accurately identify the sound
2:27
of the previously learned flies in the lab.
2:30
Our goal is to help forensic experts
2:32
in generating a large data to be able
2:34
to study and compare to criminal cases.
2:37
So next time you're cooking barbecue
2:39
and the flies can smell the meat, please remember
2:41
that all they're trying to do is preserve the time.
2:44
Thank you.
2:45
(soft music plays)