Noise and hiss reduction from MP3 tracks

ramakriya · Post by **ramakriya** » 05 Feb 2008, 03:22

I have a few mp3 tracks with lot of hiss/background noise. I tried softwares like Goldwave, wavepad etc - but none of them is really giving a good result (That might be due to the settings I am using because of my ignorance)

If anyone here has successfully reduced noise from mp3 files, please post your experience/recommendations

-Ramakriya

arunk · Post by **arunk** » 05 Feb 2008, 03:30

I always get mixed results.

In Goldwave, try attenuating freqs > 3K and it may help reduce hiss. For hum, freq < 70 may help.

3K could be aggressive. If not, try 4K or 5K.

You may not be able to take it out completely, but can lessen it.

Arun

ramakriya · Post by **ramakriya** » 05 Feb 2008, 05:18

Thanks arun

This gave a better result than what I was getting earlier!

-Ramakriya

coolkarni · Post by **coolkarni** » 05 Feb 2008, 06:42

There is another fine way.I am not familair with Goldwave or Wavepad , but these features must be avaialable here too(like Cool edit, which I use)

1.To Start with , when you start running a tape for conversion, press the Pause button for about 20-30 seconds.
2.The sound that is now appearing on the wave is the Useless part that needs to be deducted from the actual wave that you end up recording..
3.now release the pause button and complete the recording.
4.Save the wave file.
5.Now Choose -by selecting -the first 20 seconds of the wave ..
go to
Effects
Noise reduction..
Noise Reduction...
Get profile from selection.
And save Profile as "abc" file.

Now get back to the original Wave, select the whole wave and go through the
Noise Reduction steps again
Effects
Noise reduction..
Noise Reduction...
Instead of choosing "Get Profile"., now choose Load Profile.
and load the abc file that you had save earlier.
you can preview at this stage.
then Press OK.
Now you will observe that the software is deducting the original noise captured by you from the actual, wave.
Caution: Make sure you capture only the Noise element in the first 20 seconds or so by using the pause button.And not the Tambura Drone.

At the end of it you may still have noise in the track after the recording starts.
For that , you can get back to Aruns Methods.
BTW Arun
I am still looking up all buses which are plying from Velachery , to get a glimpse of You.

gobilalitha · Post by **gobilalitha** » 05 Feb 2008, 07:15

my head has started reeling After merely reading your SIMPLE? instructions. I do not know what will follow, if I try to put into operation, your advice. I am reminded of a story , probably it may differ from its original , since Japan was very often tormented by devastating earthquakes., a japanese pair sent their children to a quite country near by ..after a few days the japanese got a message ' YOUR CHILDREN ARE SENT BACK ..PL SEND THE EARTHQUAKE

vasanthakokilam · Post by **vasanthakokilam** » 05 Feb 2008, 07:23

Cool, that is quite clever. I try to use a similar noise reduction technique with Audacity as well and I could never get a good sample of the noise for it to work well. Your technique addresses that point with that initial 20 seconds of noise sample.

meena · Post by **meena** » 05 Feb 2008, 07:24

gobilalithaji

IT WILL be SIMPLE

Thats is AFTER u install the Cool Edit.
All u have to then is open the program and follow the SIMPLE

steps that Kji has provided.

vasanthakokilam · Post by **vasanthakokilam** » 05 Feb 2008, 07:29

That is quite funny, gobilalitha.

ramakriya · Post by **ramakriya** » 05 Feb 2008, 08:44

Coolkarni, your idea is good - I tried to use the noise profile, but becuase I did not have a 'pause' interval for the noise when I did the conversion, it did not work,

But for my next tape conversion, I will use this trick

-Ramakriya

arunk · Post by **arunk** » 05 Feb 2008, 22:08

coolkarni wrote:BTW Arun
I am still looking up all buses which are plying from Velachery , to get a glimpse of You.

- I know sorry! That last week before I left Chennai turned out to be too hectic - a direct side effect of "postponing social obligations" in order to attend as many concerts as possible. Btw, I plan to next December and will make sure we meet then!

Ashwin · Post by **Ashwin** » 05 Feb 2008, 22:51

Arun,

This is slightly off-topic, but I was wondering if you have ever come across techniques to eliminate or reduce excessive tambura volume in a recording. We often have this problem when recording concerts from microphones placed on stage, inevitably near an electronic tambura box. I would think its next to impossible, since the spectrum associated with the tambura would be similar to that associated with the vocal/instrumental content...any thoughts would be appreciated.

Ashwin

vasanthakokilam · Post by **vasanthakokilam** » 06 Feb 2008, 00:07

Ashwin, good problem you pose to solve. Theoretically, if the tambura has different harmonics than the rest of the instruments and voice, one should be able to attenuate just a specific (higher ) frequency sub-band. I do not know what effect it will have on the overall sound, but it something to try. Meaning, start attenuating only for frequencies > 10K, then > 5K. then > 3K etc. But then at the base frequency and a few partials above, how can one distinguish between the Sa of a tambura and the Sa of voice/violin?

arunk · Post by **arunk** » 06 Feb 2008, 03:07

Ashwin - yes I imagine this will be very difficult. The tampura is quite rich in harmonics - I imagine the base freq is in 100-200Hz range (maybe even less?) but it has many many harmonics going up to high freqs. Because of excessive sustain of sa-pa-Sa, you of course have the three sounds overlap each other (although they are all harmonically related). Also you have that "twang" effect caused by that thread (jiva?) - i dont know how that manifests in the frequency spectrum.

Just from a pure audio analysis perspective, it seems very difficult to separate it out. Anytime the singer touches on sa, pa, ga3 - some/many of his/her harmonics will coincide/mix with that of the tampura.

vk - its not just a matter of the base frequency, as it is not necessary for the base frequency to have the "highest amplitude" component. So to lower one of the "voices" from a signal, theoretically you need to identify all partials and do some sort of attentuation - that gets tough when the "other voices" in the signal happen to be at a pitch that is harmonically related with this (and so their partials mix in with this one).

It sounds difficult - but there may be some genius researchers out there who have figured effective separation (i vaguely remember coming across research that work at identifying different voices in chorus etc.)

Arun

Ashwin · Post by **Ashwin** » 06 Feb 2008, 03:29

This might be a stupid idea, but what if we sampled the same tambura under the same ambient conditions for a duration roughly equal to that of the total recording, and then subtracted? Since it is a periodic waveform, shouldnt this remove the tambura (adjusting for phase shift)? If so, then we would actually only need a single period of the tambura cycle (S-P-S-S), right? I'm only half serious - everything Ive ever been taught tells me it wont work (wont work well, at least)

Ashwin

vasanthakokilam · Post by **vasanthakokilam** » 06 Feb 2008, 03:32

Ashwin, Comb Filter is the technical word for this, I think. Not sure if it will take out only the tampura. You can compare two wave forms and say with some confidence if they are similar or if one is contained in the other, but can such a Comb filter subtract the tampura without affecting the other swaras that match the same frequency as the tampura? Just raising a question. I do not know.

Arun pointed me to a reference for last weekend reading on beat detection. From there I got to learn about how Comb Filter is used in beat detection.

http://www.owlnet.rice.edu/~elec301/Pro ... .html#comb

The technique is similar but in their algorithm, there is no subtraction. May be if you all read the rather easy to understand material here ( look around also ), may be that will give further ideas.

arunk · Post by **arunk** » 06 Feb 2008, 03:43

Ashwin - my (blind) guess is it will work to a certain extent but will not be satisfactory as it will affect the frequency components of the other "voices" in the signal. Basically what you are suggesting is sort of similar to what coolji suggested for noise. You treat the tampura as the "noise profile" which instead of "eliminating" you want to reduce a few notches.

Arun

vasanthakokilam · Post by **vasanthakokilam** » 06 Feb 2008, 05:25

Good point Arun. This should be easily testable with a sample recording. Ashwin, let us know what you find.

coolkarni · Post by **coolkarni** » 06 Feb 2008, 05:38

Oh !!!
How I Love listening to Tambura Volume in a CM Recital !!
Have'nt Heard it for ages.
And you are all trying to scale it down .

:D

balakk · Post by **balakk** » 03 Jan 2012, 14:49

Sorry for reviving a 4 year old thread, but thought of sharing some methods I use and learning from others of course.

- Audacity 1.3 beta : The noise removal algorithm has gotten reasonably better with this. The process is still the same, except the results are better now.

When I cant grab a noise profile off the song, the following method is useful:
Install Foobar2000(a music player) : http://www.foobar2000.org
Install the component foo_dsp_effect: http://www.foobar2000.org/components/vi ... dsp_effect

Goto Preferences -> DSP Manager , Add 'IIR filter' .
Click 'Configure', select 'RIAA Tape Vinyl/De-emphasis.'

The method used is described here: http://en.wikipedia.org/wiki/RIAA_equalization

I find the output is definitely better than the original, though there's some clouding happening. This only affects playback, it doesn't alter the song.(Foobar2000 is mostly a player).

gardabha_gana · Post by **gardabha_gana** » 26 Apr 2015, 11:55

Reviving an old thread - what's the latest-greatest (and relatively easy) way to eliminate noise and hiss from mp3 tracks?
Remembering gobilalitha who contributed to this thread and who is no longer with us......

mridhangam · Post by **mridhangam** » 29 Apr 2015, 14:50

What a topic !!!! coolji and aruns suggestions are very good with a chipping from Vasanthakokilam and bala. I would love to hear the latest from all of you.I do a lot of editing and I have tried attenuation techniques and have been successful in reducing echoes. The EQ settings of Adobe audition is very good to reduce echo to an agreeably reasonable levels. Also the Zoom recorder with which I record has a 4 track audio recording of which 2 seem to be echo type tracks and 2 seem to be treble tracks. It becomes easy when i edit. I can cooly delete the echo tracks and tweak the treble track to give extra bass and make it listen decent.

Don't know any technical terms like all of you.
Thank you in advance.

Nick H · Post by **Nick H** » 29 Apr 2015, 16:31

Also the Zoom recorder with which I record has a 4 track audio recording

How curious! Probably you can set it to record just two tracks [=stereo]?

mridhangam · Post by **mridhangam** » 05 May 2015, 15:40

Nick
There is an option to record stereo also but as I said if you record with 4 track you can eliminate the echo tracks. Stereo recording doesn't have this facility ..

Mannarkoil J Balaji

Nick H · Post by **Nick H** » 06 May 2015, 01:10

Very curious, Balaji. I have not encountered "echo tracks" but it is a long time since I recorded anything. I keep a mildly interested, distant watch on music technology though.

What happens when you use your device at home, recording the musicians direct, with no interference from stage and house speakers and hall acoustics?

kittappa · Post by **kittappa** » 06 May 2015, 11:51

Dear mridhangam, I would like to know the method to reduce or even eliminate echo in two recordings of Semmangudi, both studio recordings. An otherwise good recording with PMI on mridangam does not sound well with the echo. Please explain what I'll have to do.

mridhangam · Post by **mridhangam** » 07 May 2015, 15:45

Sir

I do the following to get a decent echo free version. Open the desired MP3 file or .Wav file in Audition in multi track mode. Then go to the Mixer settings and enable EQ. Then you will get a menu where you can see equalisers on the left side it will run from 20hz to about 800hz after this 1khz to 20khz will be the on the right side depending on 10 band 20 band 30 band equalisers. Even if you get a 10 Band equaliser the left 500hz will be bass sound end point 1khz will be treble starting point. if you tweak the bass sides around 125Hz 160Hz or 180Hz and slightly increase the treble at 1khz till about 3khz you can get a decent echo free recording. Mridangam thoppi seriya kaekkama pogalam for that you need to have 30 band equaliser and increase the 600-800Hz range a bit.
Generally I do this to get a decent recording.

More suggestions welcome. I may be wrong also ....

Mannarkoil J Balaji

Nick H · Post by **Nick H** » 08 May 2015, 02:06

I don't know, Balaji... I used to record with a mini-disc machine and a Sony stereo mic. The recordings were excellent, but since I stopped pretending to be a music student, I have not used it.

I find the gadgets fascinating, and that is one reason I don't record live concerts: I don't like the gadgetry taking my mind off the music, but if I was going to record, I would want it to be a good recording, so I would have to mess about. Then the messing about would not be discrete, and recorders are not always welcome in the audience.

The downside is that once-in-a-lifetime is just not enough for some concerts

rasikas.org

Noise and hiss reduction from MP3 tracks

Re: Noise and hiss reduction from MP3 tracks

Re: Noise and hiss reduction from MP3 tracks

Re: Noise and hiss reduction from MP3 tracks

Re: Noise and hiss reduction from MP3 tracks

Re: Noise and hiss reduction from MP3 tracks

Re: Noise and hiss reduction from MP3 tracks

Re: Noise and hiss reduction from MP3 tracks

Re: Noise and hiss reduction from MP3 tracks

Re: Noise and hiss reduction from MP3 tracks