alt Hacker NewsHow does this work in practice. If a microphone is up there, it's constantly listening for things right?
So how do humans here on Earth go over it to know if a sound was picked up knowing there's hours of recording?
Is it that the whole system is programmed to show a spike when sound is captured?
Replies
Listening to hours of recording doesn't even seem like a lot considering this is the only microphone we have on another planet. You would need like 4 people doing this full time, which is a drop in the bucket for a project on this scale.
Of course this is not how it's done and almost all of the recording will just be wind or noise from the rover itself, which can easily be filtered out.
If you have enough RAM, start with a ring buffer.
On interesting event: compress and transfer the relevant chunk of audio from the ring buffer back to Earth.
Interesting event trigger ideas:
1) loud sound after quiet time
2) manual timestamp request from control
3) video clip recording
4) midnight, sunrise, noon, sunset. These are mostly so you have some daily baseline.
5) science package running
6) rover moving
7) abrupt camera change
Well, there are these things called computers, and they’re really very good at this stuff. It’s not exactly rocket science (heh) to write a program to listen to an audio stream and mark and log every occurence of something else than background noise and ambient wind sounds (if Martian winds are even loud enough to make sound). Everything else that the rover has to do automatically is way more complicated.
It’s pretty likely that the entire stream of silence isn’t being stored, or sent to Earth, only the interesting parts. There isn’t any way for people to listen in real time anyway, because communications (can) only happen at specific times of the day. Every interplanetary mission works by sending a preplanned sequence of commands one day, then coming back the next day to see what the probe/rover/whatever sent back, then planning the next set of commands, and so on.
This doesn't require anything fancy. I haven't used my sound engineering qualification in 14 years and I could do it by hand. You can visually scan through the recorded waveform and look for shapes that stand out. Simple audio processing techniques like using a noise gate to shut off the volume whenever the input level is below some configured threshold can make this even easier.