0
\$\begingroup\$

I am currently doing a project that require sensitive microphone as in human ear for voice recognition - but not sure where to start.

From what i understand common microphone could only pickup sound few inches away, while our ear can hear many thing at a distance; quacking bird, boiling water in kitchen, meow cats, etc

Therefore i need to be able to pickup something like 30 feet distance, Omnidirectional, and relatively small in size (just like our ear do)

How can achieve that? I am fairly new in this area so any suggestion approach would be very appreciated!

\$\endgroup\$
1
  • \$\begingroup\$ Start by studying what is used for this problem in existing solutions: specifically multiple spatially separate microphones (often MEMS) feeding a DSP "beamforming" algorithm which attempts to identify and spatially isolate sources of speech. With the current trend towards cloud voice, the beamforming is available as hardware for terminal devices, separate from the voice recognition which (apart from unvalidated wakeword) does not run locally. The development kits come with suitable microphones already mounted at correct spacing. But part selection questions are off topic here. \$\endgroup\$
    – Chris Stratton
    Commented Jul 6, 2020 at 10:24

2 Answers 2

Reset to default
2
\$\begingroup\$

The main point here is that the microphone will pick up all sounds at any distance. The problem, as has been already discussed, is that ambient noise plus electronic noise will limit your ability to distinguish them. Our ears (ears is important since we have two which helps the process), coupled with our brains, have significant signal processing capability that allows us to recognize particular sounds even at very low levels. You need to be able to duplicate at least some of that capability if you want to mimic ears.

\$\endgroup\$
2
  • \$\begingroup\$ Means, the main reason commonly-found microphone can only pickup sound several inches away is simply because there is no "brain" that process those signal, right? Since any microphone able to pick at any distance. And what need to do is exactly amplify the signal & eliminate all the noise - then it will be closer to mimic ear hearing capability? \$\endgroup\$
    – gijoe
    Commented Jul 6, 2020 at 21:30
  • \$\begingroup\$ You should do some more research on the characteristics of the human ear in terms of frequency response, sensitivity, directivity, linearity, etc. It is not easy to emulate that behavior with just a microphone and an amplifier. For example, the frequency response of the human ear is a function of the sound level. At low levels, the ear is not sensitive to very high or very low frequency sounds. This improves as the level increases. \$\endgroup\$
    – Barry
    Commented Jul 6, 2020 at 21:57
2
\$\begingroup\$

A microphone can pick up sounds of any level. However, it also generates noise, as does the amplifier following it. Very quiet sounds will be indistinguishable able the noise.

You therefore need to concentrate on the Signal to Noise ratio (SNR) of your microphone + amplifier system. I emphasise system, as there are many sources of noise, and you need to concentrate on reducing the dominant source for any given setup.

For a purely passive electrodynamic microphone, like a moving ribbon or moving coil, the electrical noise of the amplifier that follows it will dominate the inherent resistive microphone noise. You will also have wind noise, and environmental noise conducted through the microphone mounting, which will almost certainly dominate any quantum noise processes.

Different microphones will need different amplifiers to optimise their noise depending on output impedance. A moving ribbon mike has a very low output impedance, and usually uses a transformer to raise the impedance before amplification. A moving coil mike can be fed directly into a low impedance bipolar input amplifier.

If you buy a high output active microphone, then there's little you can do electrically to improve the noise from what it provides.

As Chris Stratton points out, multiple microphones can be used to create a beam pointing at the source. It's still possible to get omni behaviour by create the beam on the fly using clever DSP.

\$\endgroup\$
3
  • \$\begingroup\$ The last point is only correct as long as there's a single microphone. Today's usual voice recognition and better videoconferencing systems use multiple so that they can perform beamforming to isolate the source of speech and reduce ambient noise. \$\endgroup\$
    – Chris Stratton
    Commented Jul 6, 2020 at 10:31
  • \$\begingroup\$ A microphone can pick up sounds of any level Will that means mic product selection is not relevant question when it comes in getting the mic able to pickup sound in far distance? (i.e this INMP441) invensense.tdk.com/wp-content/uploads/2015/02/INMP441.pdf - or are you suggesting to create the mic from scratch with that 'moving ribbon' properties? \$\endgroup\$
    – gijoe
    Commented Jul 6, 2020 at 21:24
  • \$\begingroup\$ Creating from scratch is a mug's game. Choosing a low noise high sensitivity mic is important, ie one that can be used in a system with good SNR. Just low noise isn't important by itself, a dead short has zero noise. Just high sensitivity is not enough if the mic is very noisy. You have to consider the whole system. If you can sacrifice omnidirectional for directional, then a whole new world of low noise opens up as you can combine mics. Recommend you buy a cheap mic, borrow an expensive one, and do some experiments. \$\endgroup\$
    – Neil_UK
    Commented Jul 7, 2020 at 5:44

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