removing loud audio hum from voice recording

I'm working on a magazine article, and I used a Sandisk Sansa to
record a ~45 minute audio interview to use as my source material. Now
that I get around to reviewing it, it sounds ok for the first second or
two, then an audio hum overwhelms the audio to the point where I can
just barely make out the speaking, but not well enough to get anything
out of it, no matter how long I spend straining. I originally thought
this was an issue with the audio out headphone jack on the Sansa
Express, but the problem persists even after I transfer the file to my PC.

The Sansa Express records Microsoft ADPCM, 89 kbps bitrate, 4 bit
sample size, in mono, and a sample rate of 22 kHz. It appears to be
saved in a *.wav package.

Anyway, when I play the file with VLC Media Player, and I choose the
oscilloscope visualization, it shows two distinct wave patterns. I'm
wondering if one of them is the audio hum, and if there might be some
way to use software to edit the adpcm file so that the audio hum is
removed, or at least lessened, and I can then listen to the recording
and use it for my article.

Anyone have experience with something like this? Thanks!

wrote:
> On Sat, 07 Feb 2009 22:40:21 -0500, OhioGuy <> wrote:
>
>> I'm working on a magazine article, and I used a Sandisk Sansa to
>> record a ~45 minute audio interview to use as my source material. Now
>> that I get around to reviewing it, it sounds ok for the first second or
>> two, then an audio hum overwhelms the audio to the point where I can
>> just barely make out the speaking, but not well enough to get anything
>> out of it, no matter how long I spend straining. I originally thought
>> this was an issue with the audio out headphone jack on the Sansa
>> Express, but the problem persists even after I transfer the file to my PC.
>>
>> The Sansa Express records Microsoft ADPCM, 89 kbps bitrate, 4 bit
>> sample size, in mono, and a sample rate of 22 kHz. It appears to be
>> saved in a *.wav package.
>>
>> Anyway, when I play the file with VLC Media Player, and I choose the
>> oscilloscope visualization, it shows two distinct wave patterns. I'm
>> wondering if one of them is the audio hum, and if there might be some
>> way to use software to edit the adpcm file so that the audio hum is
>> removed, or at least lessened, and I can then listen to the recording
>> and use it for my article.
>>
>> Anyone have experience with something like this? Thanks!
>
> A good audio editor like sound forge can filter out a single
> frequency (like 60hz). You may lose a little of the content but the
> hum will be gone.
>

Now I'm not sure it is a specific frequency. I looked at the
waveform, and it looks like little 'spikes' going down into negative
territory - about 100 per second. However, when I play it through my
stereo and use the graphic equalizer, a bit more of the noise disappears
when I move the upper frequency sliders down. When I zoom in even more
to the wave, I see that the noise is a pulse that lasts 3 ten
thousandths of a second, and begins every hundredth of a second. It
looks like a v, with the bottom of the trough going down to about -.18
from zero.

No matter what I do, though, it appears that since the hum is 2 to 3
times as loud as the people speaking, I am having a really tough time
making out what they say. The annoying thing is that the Sansa Express
recorder will not duplicate this problem - it appears that the thing
only had this problem when I was sitting in their office near the
computer and a bunch of other electrical equipment. Hence, I'm thinking
it might have been RF interference of some kind. Every recording I do
now is quite clear.

I used Audacity to apply a high pass filter, then tried a low pass
filter. No help.

I think next time I'm just going to take a backup recorder, and have
two of them running, just in case.

On Feb 8, 9:31*pm, OhioGuy <n...@none.net> wrote:
> gfretw...@aol.com wrote:
>> On Sat, 07 Feb 2009 22:40:21 -0500, OhioGuy <n...@none.net> wrote:
>>> * I'm working on a magazine article, and I used a Sandisk Sansa to
>>> record a ~45 minute audio interview to use as my source material. *Now
>>> that I get around to reviewing it, it sounds ok for the first second or
>>> two, then an audio hum overwhelms the audio to the point where I can
>>> just barely make out the speaking, but not well enough to get anything
>>> out of it, no matter how long I spend straining. *I originally thought
>>> this was an issue with the audio out headphone jack on the Sansa
>>> Express, but the problem persists even after I transfer the file to my PC.
>
> >> * The Sansa Express records Microsoft ADPCM, 89 kbps bitrate, 4 bit
> >> sample size, in mono, and a sample rate of 22 kHz. *It appears to be
> >> saved in a *.wav package.
....
> * *Now I'm not sure it is a specific frequency. *I looked at the
> waveform, and it looks like little 'spikes' going down into negative
> territory - about 100 per second. *However, when I play it through my
> stereo and use the graphic equalizer, a bit more of the noise disappears
> when I move the upper frequency sliders down. *When I zoom in even more
> to the wave, I see that the noise is a pulse that lasts 3 ten
> thousandths of a second, and begins every hundredth of a second. *It
> looks like a v, with the bottom of the trough going down to about -.18
> from zero.
>
> * *No matter what I do, though, it appears that since the hum is 2 to3
> times as loud as the people speaking, I am having a really tough time
> making out what they say.
....

If the interference is as consistent as you describe and you wanted to
email me with an attachment containing the first minute of your
original unmodified recording in .wav format then I would take a few
minutes and apply a chopper to try to slice away the interfering
signal. I would reply by email with the processed minute of .wav
file. If the results were good enough then perhaps we could try the
chopper on your whole file.

Or if you have the right tools you could try writing up your own
signal processor snippet to do this. From your description I'd
suggest your code look for about .3ms wide negative going pulses,
remove those, and fill the gap with a straight line between the two
endpoints where you chopped out the pulse.

No obligation, no credit, usually I learn something and that is enough
for me.
(no really)

And next time I'd suggest doing a quick sound check on at least one
recorder to confirm you have an acceptable recording under the
circumstances.