Thread: [Sv1-devel] What is the data being exported?
Brought to you by:
cannam
From: Paul C. <che...@gm...> - 2019-11-29 01:36:28
|
Hi folks, I'm new to this project and excited to learn about what it does. One quick question - what is the csv data exported by: File -> Export Audio to Data File I wanted to export peak frequency spectogram but it doesn't look like it. Thanks for comments Paul |
From: Rainer J. S. <Rai...@mo...> - 2019-11-29 15:56:57
|
Hi Paul, "Export Audio to Data File" does not save the Spectogram or another analysis, but the original audio data. I'm not sure for whom this feature was created as it results in very, very big text files with exactly the same information as in the previously loaded binary .wav file (but if you export ca. 1 sec., import it in Excel and make a line chart there, you can see the waveform). Depending on your needs you could be happy with File | Export SVG File... (as SVG contains vector information in plain text). I do not know a way to save the data from SV's FFT analysis ("Export Annotation Layer" does not seem to be implemented for Spectogram Layers/Panes). As SV and most Vamp Plugins are open source, you could implement something for this purpose or read the fft data from the .sv file. But if you want to re-use FFT data in mathematical/statistic software like R(Studio), it is easier to calculate it directly there; e.g. for R[Studio] you could use the packages tuneR and signal, have a look at https://hansenjohnson.org/post/spectrograms-in-r/ (it's for analyzing whales, but works with Wagner and The Who too). Best regards, Rainer >>> Paul Chen <che...@gm...> 29.11.2019 02:36 >>> Hi folks, I'm new to this project and excited to learn about what it does. One quick question - what is the csv data exported by: File -> Export Audio to Data File I wanted to export peak frequency spectogram but it doesn't look like it. Thanks for comments Paul |
From: Chris C. <ca...@al...> - 2019-12-05 10:43:24
|
Hi Paul, Rainer - On Fri, 29 Nov 2019, at 15:25, Rainer Josef SCHWOB wrote: > "Export Audio to Data File" does not save the Spectogram or another > analysis, but the original audio data. I'm not sure for whom this > feature was created as it results in very, very big text files with > exactly the same information as in the previously loaded binary .wav > file (but if you export ca. 1 sec., import it in Excel and make a line > chart there, you can see the waveform). As your example suggests, it's mostly useful for exporting very short selections for analysis in some non-audio tool, rather than exporting the whole of a long audio file. There is currently no way to export the contents of a spectrogram layer. Depending on the type of spectrogram, you may be able to run it in a plugin and then export the contents of the 3d plot layer that shows the plugin output. This is undoubtedly a feature that ought to be there. There are a couple of reasons why it wasn't added at the same time as exporting other kinds of layer. One is tediously technical - the export function exports whatever internal model the layer is formally a view of, but the spectrogram layer is formally another view of the waveform data (it just uses another model for conversion internally) so a naive implementation would end up exporting time-domain data rather than spectrograms! Another line of thought was just that spectrogram data may be rather large to be exported as text. But that was before the ability to export plain audio as text was added, and that's of a similar size. And of course any plugin can produce equally bulky output that can be exported. Chris |
From: Paul C. <che...@gm...> - 2019-12-08 10:44:44
|
Good points! Glad to learn about the history of this great project. I break up the sounds sampled from a metronome. Each sound wave within that time period is not big so I can extract peak frequencies of each sample using the spectrum UI. From them I can analyze the music notes being played from each sample sound. All are done manually. So I'm hoping to find a way to automate the manual steps. Now it looks like I need to hack the code to make that happen? Thanks for comments On Thu, Dec 5, 2019 at 2:43 AM Chris Cannam <ca...@al...> wrote: > Hi Paul, Rainer - > > On Fri, 29 Nov 2019, at 15:25, Rainer Josef SCHWOB wrote: > > "Export Audio to Data File" does not save the Spectogram or another > > analysis, but the original audio data. I'm not sure for whom this > > feature was created as it results in very, very big text files with > > exactly the same information as in the previously loaded binary .wav > > file (but if you export ca. 1 sec., import it in Excel and make a line > > chart there, you can see the waveform). > > As your example suggests, it's mostly useful for exporting very short > selections for analysis in some non-audio tool, rather than exporting the > whole of a long audio file. > > There is currently no way to export the contents of a spectrogram layer. > Depending on the type of spectrogram, you may be able to run it in a plugin > and then export the contents of the 3d plot layer that shows the plugin > output. > > This is undoubtedly a feature that ought to be there. There are a couple > of reasons why it wasn't added at the same time as exporting other kinds of > layer. > > One is tediously technical - the export function exports whatever internal > model the layer is formally a view of, but the spectrogram layer is > formally another view of the waveform data (it just uses another model for > conversion internally) so a naive implementation would end up exporting > time-domain data rather than spectrograms! > > Another line of thought was just that spectrogram data may be rather large > to be exported as text. But that was before the ability to export plain > audio as text was added, and that's of a similar size. And of course any > plugin can produce equally bulky output that can be exported. > > > Chris > > > _______________________________________________ > Sv1-devel mailing list > Sv1...@li... > https://lists.sourceforge.net/lists/listinfo/sv1-devel > |