Which software can do pattern matching on waveforms?

[wiseleo]

I can't quite figure out what this would be called.

I recorded a video with lousy built-in microphone on my camcorder. I also recorded high quality sound using my Zoom H2 recorder.

Zoom H2 was set to record continuously while my camcorder was only used to record less than 40 minutes of footage. This means I have 6 hours of audio, and 40 minutes of video..

What I'd like to do is to find a way to basically find the camcorder's soundtrack signature in my Zoom H2 wav file. By that I mean highlight a piece of the waveform and run a "sounds like" search for it in another file.

In essence, right now I have to listen to the soundtrack on my video recording and then try to find that soundtrack manually in my high quality wav file. I'd much prefer that software do it for me . Then I would simply replace the soundtrack with my high quality track.

Since Audacity can do noise reduction, I think it should be capable of doing this as well.

Questions for the forum:

1. What is the professional term for this? So I can run some searches on correct keywords.
2. Do you know of any software that can do this?
3. Are there any plugins for Audacity that can do this?

Thank you!

[Kayembee]

> I can't quite figure out what this would be called.
>
> I recorded a video with lousy built-in microphone on my
> camcorder. I also recorded high quality sound using my Zoom H2
> recorder.
>
> Zoom H2 was set to record continuously while my camcorder was
> only used to record less than 40 minutes of footage. This means I
> have 6 hours of audio, and 40 minutes of video..
>
> What I'd like to do is to find a way to basically find the
> camcorder's soundtrack signature in my Zoom H2 wav file. By that
> I mean highlight a piece of the waveform and run a "sounds like"
> search for it in another file.
>
> In essence, right now I have to listen to the soundtrack on my
> video recording and then try to find that soundtrack manually in
> my high quality wav file. I'd much prefer that software do it for
> me wink.gif. Then I would simply replace the soundtrack with my
> high quality track.
>
> Since Audacity can do noise reduction, I think it should be
> capable of doing this as well.
>
> Questions for the forum:
>
> 1. What is the professional term for this? So I can run some
> searches on correct keywords.

It's usually called signal processing, digital signal processing
(DSP), signal analysis.

> 2. Do you know of any software that can do this?

Not specifically, but the general context for this sort of thing
is bigger then audio, it's used in any kind of digital
collection of raw data. For example, I used to write software to
do similar things to digitized physiological data, heartbeat,
respiration, stuff like that.

In this context, signal analysis packages seem to get named
"Lab"-something. I remember using one called Labview on PCs. Very
powerful and general-purpose, not cheap but not utterly
outrageous, still available I think.

> 3. Are there any plugins for Audacity that can do this?

This is probably a better way to go; a Labview-type package would
be killing flies with a shotgun. Were I you I'd use Audacity to
Z-score ("normalize") each track, then notch filter them to the
main audible frequencies, then just do an eyeball-compare of the
two graphs to find the temporal overlap, and cut the unprocessed
higher-fi track where indicated.

Or if you were ambitious, you could write your own comparison
routine . Shouldn't be that hard, once the signals are
Z-scored and notch-filtered, and their sample rates matched.

[redwudz]

My first question is do you plan to do this each and every time you use the camera and audio recorder? If so I can't imagine why. It just sounds like you have created an incredible amount of unnecessary bother.

If it's just this once with 6 hours of audio, just find the track by listening and don't ever do it that way again. If you must, use a key sound like a 400Hz tone before you start the audio recording so you can locate it easier. That's why they used a 'clapper' with movie filming to show the beginning of a audio track on a tape recording made during filming. That sound, along with the view of the clapper on film is used to sync the video and audio together and keep track of which scene is being filmed. It's a very simple method that has been widely used for many years.



And welcome to our forums.

[Bob-S]

Changing Video Sound

You can use also video editing software to replace your sound. I have used both Premier and Ulead. They can show the soundtrack separately from the video track. Bring in both the video with sound plus the better sound track into the editor. Drag both down to the editing timeline. Then cut the beginning of the better sound to match the movie. Also delete or cut of the excess sound at the end. This may take some trial and error to get the improved sound line to match the movie.

Once you have the new sound lined up or synchronized, then delete the bad sound line and move the good sound line up to the movie sound line track.

Now you export your movie with the better sound to a file DVD etc.

[wiseleo]

Thanks, very informative.

To answer the "why" question. I want to appear amateur when I film my scenes.

I think I can an electronic clap with a remotely controlled device next to my mic relatively discreetly.

I'll do more research.

> I can't quite figure out what this would be called.
>
> I recorded a video with lousy built-in microphone on my
> camcorder. I also recorded high quality sound using my Zoom H2
> recorder.
>
> Zoom H2 was set to record continuously while my camcorder was
> only used to record less than 40 minutes of footage. This means I
> have 6 hours of audio, and 40 minutes of video..
>
> What I'd like to do is to find a way to basically find the
> camcorder's soundtrack signature in my Zoom H2 wav file. By that
> I mean highlight a piece of the waveform and run a "sounds like"
> search for it in another file.
>
> In essence, right now I have to listen to the soundtrack on my
> video recording and then try to find that soundtrack manually in
> my high quality wav file. I'd much prefer that software do it for
> me wink.gif. Then I would simply replace the soundtrack with my
> high quality track.
>
> Since Audacity can do noise reduction, I think it should be
> capable of doing this as well.
>
> Questions for the forum:
>
> 1. What is the professional term for this? So I can run some
> searches on correct keywords.

It's usually called signal processing, digital signal processing
(DSP), signal analysis.

> 2. Do you know of any software that can do this?

Not specifically, but the general context for this sort of thing
is bigger then audio, it's used in any kind of digital
collection of raw data. For example, I used to write software to
do similar things to digitized physiological data, heartbeat,
respiration, stuff like that.

In this context, signal analysis packages seem to get named
"Lab"-something. I remember using one called Labview on PCs. Very
powerful and general-purpose, not cheap but not utterly
outrageous, still available I think.

> 3. Are there any plugins for Audacity that can do this?

This is probably a better way to go; a Labview-type package would
be killing flies with a shotgun. Were I you I'd use Audacity to
Z-score ("normalize") each track, then notch filter them to the
main audible frequencies, then just do an eyeball-compare of the
two graphs to find the temporal overlap, and cut the unprocessed
higher-fi track where indicated.

Or if you were ambitious, you could write your own comparison
routine . Shouldn't be that hard, once the signals are
Z-scored and notch-filtered, and their sample rates matched.

[olyteddy]

Audio pattern matching does exist. WinAmp can retrieve ID3 tag data from GraceNote's data base based on the 'musical signature' of the particular song. While this isn't exactly what you want, perhaps in this vast interweb thingy there are some articles on how GraceNote works.

[wiseleo]

Audio pattern matching does exist. WinAmp can retrieve ID3 tag data from GraceNote's data base based on the 'musical signature' of the particular song. While this isn't exactly what you want, perhaps in this vast interweb thingy there are some articles on how GraceNote works.

Oly - interesting thought. My music is recorded in a club environment, but should be recognizable by the commercial services.

Also, it turns out that my recorder (Zoom H2) is actually equipped with ability to mark tracks digitally that can be deciphered by a BWF (broadcast wave file format) editor. Even though my files are saved as WAV, they are actually also BWFs. Audacity 1.3 beta can handle BWFs but not do much with them from what I can tell. I am going to try what Vegas 8 Pro can do with them. Another option is BWF-Widget Pro.

Now, that is damn cool!!! I can actually also record the real timecode as well onto an audio channel, but that's hazy.

Yep. I read the manuals. Last weekend was a test shoot. From this point forward it's live work.

I just need to hit the pause/play button, and it'll insert a digital mark in my recording. Assuming that I can keep my camcorder and recorder clock in reasonable proximity, I should be able to do what I want very easily.

I think I'll just do end of scene marking on my recorder and it should make my life significantly easier and keep the stealth appeal that I prefer.

[jmigsanchez]

Try this Voice Matching software:
http://www.voiceconverter.net

Regards

JM Sanchez