Removing white noise from audio

holly-louisesavage wrote on 7/12/2015, 11:06 AM

I made a recording on my iPod device using an app called "voice recorder" but the background noise (white noise) means that the speech from far away is very hard to hear. I have tried to give it a go using the audio cleaning lab but unfortunately as I don't know what I'm doing I am struggling. Would anyone be able to clean this up for me so I can hear the speech?

Comments

johnebaker wrote on 7/12/2015, 1:34 PM

Can you upload a section of the audio recording to somewhere like Dropbox and post a link to it here so we can have a look at it to see if cleaning is possible.

John EB

holly-louisesavage wrote on 7/12/2015, 3:16 PM

Thanks johnebaker! It's taken me a while to figure out how to do even that, as you can probably tell, I am not good with computers! https://www.dropbox.com/s/qaf1kqxazag52u4/Audio_07_06_2015_18_23_27_1.m4a?dl=0

johnebaker wrote on 7/13/2015, 3:53 AM

Thanks for the audio clip - unfortunately the noise level is too high to extract the voice without introducing unwanted side effects ie warbling and losing some of the speech.

John EB

browj2 wrote on 7/13/2015, 2:00 PM

Hi,

You didn't say which version of ACL that you are using. I am using the advanced version, Audio & Music Lab 2014 Premium, so I probably have more tools available than you. I also have a plugin called MDynamicEQ from Melda.

The mic is way too far away from the source to pick up the sound properly. Here is the spectral image at about 20:32. You can see that the voice volume is way down, the ambient noise is much louder, and there is a spike where a door closed of something.

I normalized the audio and tried some EQ reduction of frequencies around the voice. I can reduce the hissing a bit, but I can't bring up the voices to be very clear.

Here is the spectral display of what you recorded.

Here is a wave section just before a loud noise, normalized by itself.

Audio clip normalized section, without effects

MDynamicEQ display - I could have played with this more, but stopped here. Diminishing returns.

Same clip normalized, with MDynamicEQ

Here is the SoundCloner display.

Same clip normalized, with Sound Cloner plus MDynamicEQ

Magix 6 band EQ by itself.

Here is one with combined Sound Cloner, Magix 6 band EQ, plus the same MDynamicEQ.

I also isolated a section of voice from shocks, like the door closing and normalized. This gave me better voice. Then I played around with some EQ in Melda MDynamicsEQ to try to get the hissing down and keep the voice. It is quite difficult but I somewhat succeeded.

However, again, the voices are too far from the mic to be able to do much to make them more audible.

To at least hear them better, cut out the door closing or other loud noises, then normalize, or just normalize in sections as a first step. After that, it becomes more difficult.