Making a USB Voice Recording Better

Could use some help boosting the WANTED sound levels. Working with a USB recorder that saves in wav. I have a conversation with the USB recorder on my neck so my voice (mans) but a woman walks to me and away… The walk away voice gets too soft to hear … and a plane/jet noise flying very high above is also heard at this same time… Can hear the womans voice some of the words are louder and can be heard but faint.

I also believe this is a mono recording on the USB Recorder and can stay mono also…

What is the correct order of work to take out the noise/plane and to boost the walking away voice…
Below are some of the effects / features I played with … worked some but need better results and guidance in the correct effects/software to get the best results.

Equalizer/Plot Spectrum
I found from playing with Audacity the plane noise can be lowered using the equlizer and lowering the DB on the lower freq, first five or so, on the grafic equalizer… I also used Analyse/Plot Spectrum to get a idea of the voice freq for Me/Man & Woman in the video… hard to tell but looks like voice was in the 500 hz range and hers was in the 600hz range or those were the two spikes… Her voice is bit raspy and deep for a woman. I bumped up the 600hz range 3 (bars) to help boost the womans voice… This seemed to help with the plane noise esp.

Also used AMPLIFY and could see the soft/db levels grow with each DB boost from Amplify… I used .5, .7, 1 db boost levels and a clip also set… This seemed to help also but not the grand results I was hoping with.

Was able to use the audacity to cut audio from two long files, one close to hour and the other half hour… and could save them as wav and mp3 files so pretty cool… The wav file is 1min 15 sec and the area of voice to boost and clean up is 30 seconds since I cliped it from the 30 min file.

Are there other threads and write ups that cover what I am trying to do???
What effects would you do first and go about this task???

Seems my task will need more detail needed that the threads I read thru on the first two pages of threads…
I also played with NOISE, Low Pass, High Pass.
If you can point me in the right direction, the right order of tasks, and the correct filter/software packages to use…

Window 7, svc pack 1
4 megs ram
Installed from .exe installer, and have MP3 saver modual and it works
Audacity 2.0

thanks from MN

There is no “one way” to approach this but the way that I would probably do it:

  1. Normalize to -3 dB
  2. Equalizer/Plot Spectrum. Cut frequencies below 70 Hz severely to remove low frequency noise. Cut frequencies below about 120 Hz gently to roll off the lower tones of the man’s voice.
  3. Normalize to 0 dB
  4. Apply the Compressor effect to even out the levels. The settings will need some fiddling for best results.
  5. Listen carefully and if necessary tweak with the Equalizer.
  6. Normalize to -1 dB if exporting as WAV, -3 dB if exporting as MP3.
  7. Export.
  • The Noise Removal effect is unlikely to be useful for removing the aircraft noise.
  • When using the Amplify effect, “Allow Clipping” should not normally be selected.

Thanks Steve,

That worked better than my first few attempts…
Voice recorder is better…

Any suggestions with the COMPRESSOR and what effect to try first and the order of settings to work best… I started with ratio and had it up to 9.5 range before the noise was getting bumped up too much…

I did go back to equalizer and lowered the plane and car low freq noise after the compressor…

This was the best attempt yet, but feel it can be better yet…

I have read in the manual about the -1dB normalizing recommendation BUT not -3 for mp3’s. Why do you recommend -3 per mp3’s?

MP3 is an inexact (lossy) format. If you normalize too close to 0 dB then it’s likely that after encoding there will be some clipping. Normalizing to -3 dB allows a little extra headroom that should be adequate for all but the most extreme cases.

I have read in the manual about the -1dB normalizing recommendation BUT not -3 for mp3’s. Why do you recommend -3 per mp3’s?

The -1 dB recommendation is related to “inter-sample overs”, where the re-constructed waveform can go higher than 0dB. There is no issue in the digital domain… There is only a potential issue with the analog waveform reconstruction. Personally, I’m not convinced that DACs have trouble going over 0dB on the analog side. Nor am I convinced that I could hear the tiny amount of clipping if it happens…

MP3 is lossy compression, so what you get out is not exactly what you put in. The peaks on a reconstructed MP3 often go higher than the original peaks. (Some MP3 peaks might be lower than the original… The wave shape is slightly different.) It’s my understanding that MP3 uses floating-point internally and is not limited to 0dB. And, some decoders can handle levels above 0dB without clipping. (Sometimes people will open an MP3 in a DAW or audio editor, and they are surprised that it goes over 0dB.) But the data sent to the DAC is in integer format, and it can be clipped. Again, if there is occasional clipping I’m not convinced that I can hear it. If you are worried about defects, especially defects that you can’t hear, you probably shouldn’t be using MP3 or any other lossy format anyway… We know MP3 changes the data, but it doesn’t always change the sound.

The guys that do ABX testing to check the quality of MP3 encoding never mention peak-level or clipping as a factor in encoding quality. If there are any audible artifacts (at high bitrates,), they are usually related to transients.

So when I normalize, I normalize to 0dB and I don’t worry about it. 16-bit PCM can go from -32,768 to +32,767 (0dB) and I am not worried about going exactly that high, as long as I don’t “try” to go over.

If a wave is finished to 0 normalization/amplification with no clipping showing (i.e., brought up as high as you can get it) [exported] and then reimported to a new audacity window and still shows no clipping does that mean it’s a-okay or is it still recommended that it be reduced to -1dB (below clipping) due to unpredictability of other players or some other factor(s)?

It’s not uncommon for CD players to clip just a little below 0 dB, in fact one test report that I read tested about a dozen different (good quality) CD players, and only one of them could actually go all the way up to 0 dBFS (the others were pretty close to 0 dBFS but not quite there).
The way I look at it, slightly below clipping level is better than slightly above clipping level.

And -3dB is normally plenty loud enough for most folk - when playing back you can always turn the amp up to 11 :slight_smile:


INTERESTING DVDdoug – from what I can decipher!! :unamused: