3
$\begingroup$

What are the different kinds of noise in speech signal? How can I simulate the same in Matlab for adding to a clean speech signal?

Improve this question
$\endgroup$
2
  • 1
    $\begingroup$ Are you talking about noise created by the body, channel or recording apparatus? $\endgroup$
    – user42
    Commented Dec 27, 2011 at 9:22
  • $\begingroup$ all kinds of noise possible in a speech signal like colored noises, etc $\endgroup$
    – K V Vijay Girish
    Commented Dec 27, 2011 at 13:08

3 Answers 3

Reset to default
2
$\begingroup$

Types of noises which you could try, that come to my mind at this time are:

  • White noise,

  • pink noise,

  • impulsive noise,

  • babble(multiple persons speaking in the background)

You can generate white(awgn)/pink noise(pass the white noise through a filter) in matlab. You could record impulsive noises(pen clicks, knocks on the table, etc), and babble(record yourself and or others multiple times and overlap the recordings)

Improve this answer
$\endgroup$
3
  • $\begingroup$ You cann't quite get proper pink noise by passing the white noise through filter! $\endgroup$
    – Dipan Mehta
    Commented Apr 17, 2012 at 9:24
  • $\begingroup$ @DipanMehta I wasn't aware of this. Could you explain why not? $\endgroup$
    – TwoSan
    Commented May 20, 2012 at 13:01
  • $\begingroup$ pink noise is essentially $1/{f^k}$ - while it well diminishes by about x dB per octave in log scale, where close to zero, it would approaches infinity near zero. In real life what we generally tend to refer to as pink noise, such as flicker are not actually true 1/f yet are dominant at some key frequency close to DC in someway. $\endgroup$
    – Dipan Mehta
    Commented May 21, 2012 at 6:07
2
$\begingroup$

Here are a bunch of noise samples used by a military speech research unit:

http://spib.rice.edu/spib/select_noise.html

They are free to download, and are available in both wav and Matlab binary format. If your speech material is not recorded at the same sample rate as the noise samples, it is important to resample either the speech, or noise, or both before summing the signals.

Note: pink, white, and babble noise samples are included as well as some military specific types of noise that will probably be of less interest.

Improve this answer
$\endgroup$
2
  • 1
    $\begingroup$ dead link in October 2013. Would you know of noise libraries with more kinds of noise ? $\endgroup$
    – denis
    Commented Oct 25, 2013 at 9:41
  • $\begingroup$ @denis alternative link spib.linse.ufsc.br/noise.html $\endgroup$
    – user13107
    Commented Mar 18, 2016 at 5:00
0
$\begingroup$

That depends a bit if you are dealing with a carefully made studio recording or someon yelling into a cell phone mic while walking down a New York street 5pm.

It typically is a good idea to follow the signal path and see what you may incur. There is

  1. Acoustical noise: Heating or air conditions, traffic noise, competing sound sources like music, babble, machine or work noises, birds, plane in the air, car cabinnoise and many, many, many more. It really depends on where you are and what's happening
  2. Microphone related noise: microphone handling, wind or DC air flow, proximity effect,"pop" noise. These are often a function of the type of microphone (electret vs. dynamic, omni vs. cardoid etc.).
  3. Electrical noise: Self noise of the microphone capsule, Pre-amp noise, Pre-amp distortion or clipping, electromagnetically induced or radiated noise, power supply cross talk, hum, the dreaded "ground loop", etc.

Not all these noises are applicable in all cases so an application specific analysis would be useful.

Improve this answer
$\endgroup$

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.