[qwertox]

Next step is to set up a grid of ESP32s with MEMS microphones and a FFT routine, stream that data to a server and apply some neural network to get to know the real amount of flies by tracking them individually. Then compare this to the model, and keep improving it until it fits.

[mLuby]

I want a mosquito-killing turret [1] that you could put in your bedroom and sleep in peace.

Turns out one exists [2] (in theory—no commercial product a decade later), but the one I was thinking of would use microphones rather than cameras [3] to detect the bloody suckers. Thing is, I can't find any prior research or even hints that people have considered this approach. (The closest I can find is an app [4] that detects a mosquito from ≤10cm away.) Here are some of my guesses for why it's not been done:

    A. Maybe it's too difficult to distinguish mosquito buzz/whine from background noise, even when filtering in only known mosquito frequencies (~400Hz [5]).
    B. Maybe it's too slow to sweep a room with a directional mic that's focused enough to detect mosquito sounds? (Say the turret in one ceiling corner of a 12x12x8ft room.)
    C. Maybe directional mics don't work well for all sound in their cone and so need to be able to "focus" at different distances (by dynamically changing the parabolic dish's shape), thereby increasing the room sweep time?
    D. Maybe the parabolic dish needs to be so big that it just looks ridiculous?
    E. Maybe all of this is possible but the type of microphone is so expensive that it's worth pursuing commercially?

I'm curious what you all think about this. It'd be an interesting thing to try making—who hasn't wanted to make a death ray while simultaneously furthering the fight against humanity's deadliest enemy? [6]

See also this research paper: Detecting Insect Flight Sounds in the Field: Implications for Acoustical Counting of Mosquitoes [7] (I'm pretty sure this is not directional—they hung an attractant 15cm from the mic and I don't see any dish.)

[1]: https://www.youtube.com/watch?v=BKm8FolQ7jw

[2]: https://www.ted.com/talks/nathan_myhrvold_could_this_laser_z...

[3]: I don't think cameras are well-suited to detecting something so small. The "photonic fence" from the TED talk only kinda works by adding a white background behind the bugs so it's really easy for the camera to detect them—not that practical in a normal room.

[4]: https://www.theguardian.com/science/2018/mar/29/mosquito-ear...

[5]: https://www.youtube.com/watch?v=Rcm3rPz_Q94&t=127

[6]: https://mysteriousuniverse.org/2014/08/why-mosquitos-are-hum...

[7]: https://pdfs.semanticscholar.org/a0bf/4fcd478b6214759046a739...

[mLuby]

My acoustically-minded friend had the following to say:

> Off the top of my head, I'm guessing the problem has mostly to do with background noise, and it only working in moderately quiet environments.

> As for locating the source of a sound, there's plenty of research that's been done about that. A directional mic doing sweeps is probably not your best bet, as it would be slow and difficult to cover three dimensions. I would think using an array of microphones and using volume and time delay to identify location would be much easier. Setting them up in a triangular pyramid would, I think, allow you to triangulate a sound's position in three dimensions. The problem with this two-fold: it's still not super precise, and the mosquito's whine is a steady sound, not an impulse where it's easy to identify time-delay between locations. You'd probably need to have an incredibly high sample rate to identify strong enough variations in the sound of a mosquito to effectively use time-delay, or place them far enough apart to rely on relative volume. Maybe a set of microphones you can set up around a room, rather than a single location would work better.

> I think this could be solvable by having the laser aim towards the sound and do a rapid sweep of the area, rather than a precision strike. You could also potentially have the device hit targets that will confirm they are being hit when you are setting it up so it can calibrate to the environment that it's in under regular conditions.

> Another thing to consider is that you would also probably need temperature and humidity detection, as both of those can affect how sound travels, even over short distances.

I then did more research on the killing part of this—the laser. The product I linked above uses two lasers—a low-power one to pinpoint the target and to "listen" for its exact frequency to rule out benign insects and a high-power one to give it instant heatstroke or burn off its wings. But with computer-aimed lasers (even non-burning visible green lasers), accidental blindness is a real concern especially indoors where the beam can unexpectedly reflect off glass or other shiny things. So maybe a narrow, powerful non-coherent flashlight could act as a spotlight and that'd be good enough for a human to still kill it.

[DiggyJohnson]

All this work, to suggest a fly spotlight device is hysterical. That would be marginally helpful, and maximally frustrating. I lend the project my full support were it to ever materialize.

Thank you for this analysis.