Audacity Forum

I sampled about 3 seconds of almost white noise and the program crashes. anyone have this problem?

Exactly what did you do?
What steps do I need to take to reproduce this problem?

The wav file I opened is 3.17MB. It consists of a snoring pattern 3min 30 sec long that has snore bursts every 3 sec separated by noise intervals of 2 sec. The snore bursts are ~0.3 high and the noise is 0.008-- I select 2 sec of noise,
go to effect>noise removal. the parameters noise reduction are 24, frequency smoothing 150, attack/delay time 0.15
Pressed get noise profile. After about 2 sec the error message that Audacity has encountered a problem and has to close
The details list the program info and "Offset" 000f38c4 . It occurs to me that I should probably find out more about what the parameters are about. Do you have a place to go that discusses the noise reduction.

<<<The wav file I opened is 3.17MB. >>>

That's part of what we're looking for. Just to eliminate one pathway, pull down the six second Piano Trill from my web site...

http://www.kozco.com/tech/soundtests.html

...put that on a fresh, new Audacity timeline and try to noise reduce. It doesn't matter where in the song you get the profile from, we're after the process health, not the sound quality. If it does crash, we need to know if it dies in the noise profile step or the removal step.

Koz

OK, I put the trill into 1.3.5 and it doesn't fail in the select noise or the removal part of the program.

You have bad or damaged original songs. Audacity will try and play songs which were recorded in formats that are completely foreign to it. Sometimes it will work enough to convince you that everything is OK right up until you try to do something exotic.

Then, Blammo!

See if the songs will open up in a translation program and export as something normal like 44100, 16 bit WAV. Sometimes you need to convert songs from WAV to WAV and after you do that, they start working.

Koz

The sound file is OK but I found that depending on the settings of the three parameters one can get the program to crash or not. It seems to be dependent on the spectral profile and the sensitivity parameters. It would be nice to have the program test for situations that might cause the algorithm to fail -- like dividing by zero -- instead of just crashing. I obviously don't understand the function of the three parameters in the noise removal routine. If you would like to see the file that I've been working with, let me know how I can get it to you. It might be useful for the developer of the noise reduction routine.
J

OK, I'll bite. Can you post it, or a sample of it that crashes somewhere?

Koz

I don't have a website that I can post it on. Can you suggest another method or way for me to get it to you? I shortened it to a 184Kb file and it still bombs when I tried to remove the noise between snores. I could Email it to you.

flyfischerjim wrote:I don't have a website that I can post it on

Go to Google and type in "free file hosting", then pick one. Try and find one that allows "hot linking" so you can then just post a link top the file.

You could type "free file hosting""hot linking" (with quotes) into Google.