In the process of editing the videos recorded at NI Week I discovered that one of the cameras did not have it's mic plugged in all the way and this created a very annoying high pitch ring in the audio. This affects 6 of the videos
I'm looking for assistance in eliminating this ring from the audio. Does anyone have the time to create and post a vi that will eliminate the ring from the .wav file so we can hear the presenters voice? Multiple solutions are welcome.
here is the link
Here is an attempt by one of the LV champions. I'm looking to do a little better than this
I used Adobe Audition and made a couple of passes with noise reduction filters.
Thanks in advance for any help.
I have more experience with this from an audio editing standpoint than from LabView, but it seems like it may be salvageable given enough time or a cleaver enough program.
As far as outside software goes, one option is Adobe products. Adobe Premiere is their video tool and Adobe Audition is their audio tool. I believe you get access to similar tools, so using Premiere may make it easier to combine the fixed audio back with the video later, but there are many ways to do this. These tools are all on monthly subscriptions these days, but you can get a free one month trial to each and every Adobe program, if you want to try it. Their tools are well made and have lots of tutorials online.
Other software can be found here https://www.techradar.com/news/the-best-free-audio-editor I just downloaded the first one, Audacity which I have used before. It's a free tool that will do many of the same things.
First I would start by selecting all of the audio (Ctrl+A) and looking at the frequency spectrum using Analyze > Plot Spectrum... Images 'Low Res 1' and 'High Res 1' show this spectrum with a different number of points plotted. It's log scale so the high end looks better on one and the low end looks better on the other. Using your cursor you can get the frequency for each of these peaks in the UI on the bottom right. Some of the worst peaks become obvious, such as a huge, wide hum at 60Hz (AC power) and a possible harmonic at 180Hz. A lot of loud, narrow peaks jump out above 4000Hz. By writing down the location of these peaks, I was then able to eliminate them. To make the sample below, I used a notch filter with a Q of 1 by going to Effects > Notch Filter... on the following values: 1984 (Q of 0.5), 3995, 5984, 7485, 7999, 9985, 12001, 14985, 15999, 17993, 19997. I am not sure what an appropriate Q value is, but some research or testing might help. The human voice is mostly below 1k, so I don't think it matters much if the filters up here are too aggressive. While I was trying to only cut out narrow bands, I think I more or less did the equivalent of a low pass filter and lost almost all frequencies above 7000Hz. "Attempt 1" shows the result of this, with almost all of the annoying high tones removed but a very low hum over the top. I had less luck removing the low hum, probably in part because it overlaps with the human voice so much. I had trouble uploading the audio file so I had to zip it.
On the flip side, the sample from the champions forum did a good job of removing the low hum, but left in some high tones. Doing the same analysis I did on the raw audio file but instead to the champions file, there are a few high tones that obviously stand out and can probably be easily removed. One important thing I did notice about this file is that the remaining high tones are not constant; rather they fade in and out. So looking at the frequency analysis of the whole file might hide them, versus looking at smaller sections of the file or a frequency vs time heatmap.
Doing it this way, for me at least, is a fairly manual process of trial and error. I have only a rough sense of what certain filters should be doing (notch, high pass, low pass, etc), but I don't have a deep understanding of the effect of different parameters of these filters (Q, dB rolloff, etc). The heatmap style graphs which plot frequency analysis as a function of time are also quite useful in seeing trends in the noise.
I'll try to check back here again, since I don't think I get notifications from these forums. Best of luck!
This file has been modified, our sister company, Sound Enhancement Products, looked at this on his lunch break, I thought before he spends much more time on it I would bounce it off of you to see what you think.
The way that this was done was to try to sample the noise while the speaker was not speaking and then use that as the filter. He says that he may be able to refine the process some more.
Let me know what you think…
The file is Cleaned_up.wav in your shared drive.
Matt and Anthony Thanks for the assistance,
Using Anthony's and others suggestions I have developed a LabVIEW based solution that has worked on one of the videos so far. For those that are curious I will post the code when I have it more refined.
The audio file that I shared seems to be the hardest to get correct as the speakers voice is close to the same volume and frequency as the noise. I may need to add closed caption to this one to make it shareable.