In a way, I wish that the government had put up noise level measuring equipment in every populated area of the country, allowing me to see on a map which areas are the most quiet and peaceful over time.

Sadly, this screams "surveillance nightmare" to me, as these devices probably would be, or could act as, microphones, and thus record the actual sounds/conversations rather than just checking the noise level.

Is it possible to make devices which only measure the noise level, but cannot record/pick up actual speech or human communication? And if so, are those cheaper than the microphones, or would they have to be specifically manufactured for more money? (And thus very likely not be done.)

Is this a "thing" which exists? Or would it have to be expensively invented and then have surveillance-prone people convinced to use them instead of microphones?

  • $\begingroup$ I have a vague idea that, to achieve this, you'd want to Fourier-transform the sound, throw away the phase information and keep only the power spectrum, then either inverse-Fourier-transform back again or integrate over frequency. For really robust privacy protection, you'd probably want to do all of that acoustically, before ever transducing the signal into electrical form. $\endgroup$ – Daniel Hatton Dec 10 '20 at 12:24

How do you define a microphone? Because there many ways to detect mechanical vibrations, but all of them could fit in a more or less relaxed definition of microphone. For example, you could measure temperature variations on a noise-absorbing foam. Or vibrations on a building wall for low frequencies, etc.

Following your concern of surveillance:

For surveillance, you need several components:

  • What is the transmission rate of the device? You can put a microphone, but if the device only transmit a single value every 1 minute (< ~1kbps), hardly this can be used for surveillance.
  • What does the device transmit? Transmitting a wave form to the central, can be used for surveillance. On the other side, transmitting noise level can not be used for surveillance.
  • What frequency the microphone is able to record? A sensor could be designed to detect specific frequencies but not those of speaking persons: <100 Hz, or >5 kHz will hardly detect any voice.

Most noise level detectors use similar sensing mechanism than microphones: piezo-electric or coils sensors. However, is the electronic put on it which makes a noise sensor out of a microphone. Consequently, a basic/cheap microphone is cheaper than a noise-level sensor. But don't forget about transmitting the data, this is the complex part, not the micro.

  • $\begingroup$ " if the device only transmit a single value every 1 minute (< ~1kbps), hardly this can be used for surveillance" This reminds me highly of "Hey Google" or "Alexa" code words. The system is "listening" 24/7, but in a very limited manner until they hear these sounds. $\endgroup$ – Cort Ammon Dec 10 '20 at 3:23
  • $\begingroup$ Alexa and similars are connected through your high-speed internet all the day. So definitively they ARE able to spy you. This sentence refers to devices with low CPU capability and low bandwidth communication. $\endgroup$ – Adrian Maire Dec 10 '20 at 7:20
  • $\begingroup$ I would personally NEVER EVER install a system like those in my house (except maybe protected by air-gap, but they wont work with air-gap right?) $\endgroup$ – Adrian Maire Dec 10 '20 at 7:30
  • $\begingroup$ Well, there's always the question of someone actively making the software turn on, which is an open security question. But the specific feature I was thinking of is that they operate in a low power mode to save batteries. This mode relies on an asic to listen for the wake word sipping microwatts of power. When they hear their name, then they wake up and turn on the rest of the processing. The ability of that chip to wake the system is what reminded me of this... very low bandwidth signal: wake/no-wake $\endgroup$ – Cort Ammon Dec 10 '20 at 15:42
  • $\begingroup$ No.. forget about it! To implement neural-network to detect a specific word use already far more CPU power than "microwatts" and far more than the necessary to send sound record to a server. Either you trust them or not, but they CAN (and probably DO) spy you (intentionally or by security leak). Also, please, stop this conversation here, it is not related to the question. $\endgroup$ – Adrian Maire Dec 10 '20 at 19:13

You can measure noise level in decibels with sound level meters that give an instantaneous reading of the noise level without recording the actual audio. Prices vary, but there are even free apps that measure dB's fairly accurately. Another component of the "peaceful" requirement is quality of noise. A rain storm or ocean waves breaking at 60 dB is technically louder than a 50 dB noise of a jackhammer in the distance, but which would you rather listen to?

  • $\begingroup$ they work by calculating the noise level detected by a microphone... right? $\endgroup$ – user253751 Nov 11 '20 at 18:14
  • $\begingroup$ Some, yes. Technically you just a need a pressure transducer and something to condition the signal. $\endgroup$ – jko Nov 11 '20 at 19:17
  • $\begingroup$ A fast-responding pressure transducer in air is called a microphone $\endgroup$ – user253751 Nov 11 '20 at 23:36

IMHO, if you use a device that can detect noise level if you program it in some way then it will be possible to program it in some smart way to record conversations.

Personally, I don't think that there is a way. The only thing I would suggest is some very low capabilities device capable of noise levels, with limited memory capacity and programmed to determine the maximum noise level. Also limited network bandwidth would be good to limit the amount of data transfered. In that way, it would be difficult to make this device usable.

Apart from that I don't see how it would be possible.

PS: Over the years I've read techniques that I'd never thought possible. One of them is the following. Using wifi signals to track and identify people in another room. Ever since I read that I realised that there is not much you can do in terms of privacy, if some third party is intend on violating that.


The key will be to drastically limit the hardware capabilities of the device.

I think you're going to have to start with a microphone anyway. Normally you would buffer this signal and feed it right into an ADC, but instead you can build a hardware volume measurement circuit. Make sure the circuit has a response time slow enough that audio can't be read through it with quick sampling.

You can then connect that to a regular microcontroller ADC. The lower the specs on this the better, since that will limit the options for "clever tricks".

Lastly you'll want to publish the design and source code, so that people can verify all of this. You could invite some party like the ACLU to take a random one down and have it audited.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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