0
\$\begingroup\$

I am working on an idea these days which is related to the designing and development of a ear phone which enables us to select a single voice and boost it out of many without any noise.Which will be very much useful when we are in a crowed. The frequency of everyones voice are distinct and therefore we could use several miniature audio receivers to process the multiple voice.Therefore we could eliminate unwanted sound and noise while we are in a crowed and concentrate on single one.

Guys I need suggestiona on this topic and please do help and correct me if I was wrong.I am new to this platform.

\$\endgroup\$
3
  • 2
    \$\begingroup\$ You're making some big assumptions there. A human voice contains many harmonics at different frequencies, and the pitch of a person's voice goes up and down as they speak. \$\endgroup\$
    – Simon B
    Oct 16, 2017 at 15:47
  • 3
    \$\begingroup\$ Your idea might work if we all communicated by whistling in Morse code - and all on a constant unique note. We don't. The human voice contains a broad spectrum of frequencies that are constantly changing in frequency and amplitude. \$\endgroup\$
    – Transistor
    Oct 16, 2017 at 15:54
  • 3
    \$\begingroup\$ Quite a complex signal processing project: the best approach is not based on frequency but spatial separation. The more microphones the more spatial separation is possible. This is why some of the "virtual assitants" (eg amazon echo) have microphone arrays. \$\endgroup\$
    – pjc50
    Oct 16, 2017 at 15:59

2 Answers 2

2
\$\begingroup\$

If the background noise is really continuously equal (like a machine), it can be removed by sampling and subtracting the sample. However, multiple changing frequencies will be hard to process (if possible at all).

However, you maybe can use some ideas of noise cancellation head phones.

Also, consider using multiple microphones and based on where a person is (close/far) process the signals. This way it should be possible to increase or lower the strength of a nearby voice. But even this is far from trivial.

The best idea is probably to use a very direction microphone directed towards the person to pick up the voice from. However, if you move or he moves it might change the signal considerably.

To start with something simple: remove frequency bands which are not used much by voice (like very low/high bands). But the human voice is very dynamic, both in volume (changes) and frequency (changes).

\$\endgroup\$
1
\$\begingroup\$

If you want to use one microphone to pick up the signal, it can be a very difficult task. Depending on how much the user can change the parameters, I can give a few ideas.

You can possibly use a uni-directional microphone and find a way to pinpoint the direction of the microphone towards the person you are trying to listen to, then use a bandpass filter with a bandwidth of about 300Hz to 3kHz (typical frequency range of the human voice). That will filter out any other noises that are of higher or lower frequency that aren't useful to what you're trying to capture. Of course, if there are other people in the path or if there are other sound sources within that 300Hz to 3kHz frequency, it can be very screwy.

300Hz to 3kHz is a very wide range, and encompasses almost everyone's range of vocal frequency. If the user has access to the filter parameters, a another solution would be to increase the Q-factor to narrow the bandwidth, and also be able to change the center frequency of the bandpass filter to adjust closer to the person's frequency range.

Then there's an issue of distance. If you want to use a standard headphones microphone or the microphone on a smartphone, then you will be very limited in what you can do.

Another way to do this would be similar to how the a person can create music tracks the isolate the vocal track of a song (a.k.a. an acapella). Basically, if the record company releases a "instrumental only" version of a specific song, a person can take the version, reverse the polarity of it, and combine it with the original song that has the vocals. This will cause the instruments to cancel with destructive interference, while the voice remains.

The same idea can apply here IF you somehow have the background noise WITHOUT the vocals you're looking for. I find this unlikely, unless you plan on using the setup for a staged scene that has an expected and consistent background noise each and every time you want to record the intended person's voice.

These are just a few suggestions and ideas that are off the top of my head. Like I said, this can be a very difficult task, and it can be more or less difficult depending on your knowledge and experience in audio and digital signal processing and frequency spectrum analysis.

\$\endgroup\$

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