How to remove different voices?

[nharikrishna]

Hello all,

I have a very old audio recording of my grand-mother singing a song which was recorded on an audio cassette. Now I have converted that to digital WAV file by recording it into the computer using Canopus ADVC 110 'Audio-only' mode.

I have Sound Forge pro, and Audacity but not sure how to use most of their features.

With trial and error, I managed to remove the tape 'hiss' using Audacity.

Now the real problem: While my grand-mother was singing, there were 2 more sounds coming out all at the same time: A baby crying, and the sound of a fan.

How do I separate out those 2 voices (Babay crying, and the fan) and retain only the grand-mother's voice? I don't mind if it is Audacity or Sound Forge.

Thanks for the advise.

[redwudz]

It depends on the pitch of those interfering sounds, along with how repetitive they are.

Repetitive sounds can sometimes be lessened by using the Noise Removal filter in Audacity. Clip a sample of the sound and try a few settings. You can always restore to the original.

The baby sounds you might be able to lessen with a filter that notches out those particular frequencies.

The problem with either filter is if the desired audio shares the same frequencies. If so, it may also be filtered and it's levels lessened.

Most times for me it is trial and error for filtering until you become more familiar with the characteristics of each filter and the program that uses them.

Someone here may be able to give you more specific suggestions.

If you can post a short representative sample, about 30 seconds, of each type of interference, someone may be able to help you more.

[nharikrishna]

Thanks red, I see what you are saying.

How to determine the frequency of a particular sound (for example, baby sound), when both main and the baby sounds are audible at the same time?

I will post a representative sample however it won't be until tonight.

Regards

[redwudz]

For the baby sounds, there is undoubtedly a place on the audio with just them, such as a pause in the main voice parts. Even a very short interval might be enough to experiment on. Likely the same with the fans sounds. I would never expect miracles, but sometimes you can make a decent improvement in the audio.

[SLK001]

You probably won't have much luck removing the baby cry (without destroying that what you want to keep), but the fan shouldn't be too difficult, provided it isn't an oscillating fan (thus varying pitches). To remove the fan noise, find a piece of audio that just has the fan noise (the larger the selection, the better) and create a noise profile, then use the profile to reduce the noise.

I don't know if Audacity has the capability to reduce noise like I suggested, and I haven't used Soundforge in years. I DO know that Adobe's Audition (formerly, Cool Edit Pro) can do it - and quite well.

[jagabo]

I don't know if Audacity has the capability to reduce noise like I suggested

Yes, it does. But it's not likely to work with "fan noise" if he means the whooshing of air which will be pretty broad, spectrally -- basically a pink noise or white noise.

[PuzZLeR]

I thought the sound of a baby's crying was indeed a miracle instead of removing it.

Seriously though, I would like a sample of this if you can give us one. I'd like to also test it on GoldWave.

I'm no expert in this myself, but the best way from my point of view is to cut the video into many parts, such as the baby crying separately, and even segment it according to several pitch ranges of the fan and do each independently with experimentation, and rejoin and normalize the audio afterwards to make it constant.

This is how I did such stuff in the past - but it's a home method, not professional.

BUT, whatever you do, and use for your playback - keep the Source separately for archiving with all the extra sounds (the tape or the ripped raw WAV file).