Is there a technology which allows to record relatively quiet sounds even in a (very) loud environment? To perhaps better explain what I'm after, picture a small ballerina music box at a loud festival / next to a passing freighter train / close to a rocket launch - I want to record its music and ignore the loud environmental sound.

Actual purpose of this would be to record the (relatively) quiet echos (off walls, ground) of a short "beep" in different environments, some of them potentially very loud (much louder than the "beep" itself) - for echolocation purposes.

Note that this is not about post processing the audio after recording - I know how to approach this; I more concerned about the recording itself.

3 Answers 3


No microphone can differentiate between different types of sound, only different frequencies.
Human ears have a brain adapted to picking out one sound amongst others, but microphones cannot do this. Some element of AI processing would need to be done. If you're dealing with a pure sine at one specific frequency, that task might be easier, but there will still come a point at which you can no longer differentiate.

However, if you're thinking in echolocation terms, then you could do worse than to look how bats do this. All bats use frequencies above the human hearing range. Some at 'only' 21kHz or so, some right up at 110kHz. Their volume output can also vary between 110dBSPL & 128dBSPL. That's loud.
They can still struggle a bit in urban environments due to background noise.

From an electronic perspective, you could test how high you need to go to be well above the general frequency spread of your noise floor, then put a high pass filter on the rest.


This doesn't specifically answer your question but, based on what you asked, I think this is what you need to know.

For something like echolocation, you would be using a signal with a specific frequency (high for precision or low for long range). You don't need to block out all the other noise - you just need to look for a corresponding return of the specific frequency you sent out. If you sent out a 40kHz signal, you would be looking specifically for a 40kHz echo to return.

You could do this with filters set to remove everything above and below your chosen frequency or by choosing a frequency band that is normally clear of interference.

A good system might send out multiple frequencies as failsafe's for any signals that do get masked by other noises or they might be able to switch frequencies to overcome a persistent source of interference or conflict with other echolocation systems operating nearby.

The frequencies used are often beyond the limit of human hearing (bats use up to 100kHz) which not only helps to distinguish the return signal from environmental noise but also gives them a strong sense of the density of any objects they encounter.

Parking sensors use echolocation by ultrasound for object avoidance. Using the ultrasound range means very little environmental interference - they still work despite being mounted on a noisy car in a noisy street.

A relatively inexpensive ultrasound echolocation device can be made using off-the-shelf parts and open source code - lots of STEM kits for kids include the parts for this


You just use good microphones, good preamps and A/D converters, and your "beep" is a specific multispectral sound with good autocorrelation properties. You then do cross-correlation of your beep with the microphone signal. There is a study for the kind of sequences to use here: try looking up the terms "perfect sequence" and "almost perfect sequence". Those tend to be binary or close, meaning that cross-correlation with them is fast to calculate.

Since they are pseudorandom, cross-correlation with them tends to strongly attenuate anything not originating from echos of themselves.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.