Opening and re-saving audio files

fiji
imagej
format
converters

#1

Hi there,

Something of a long shot, and more of a creative than a scientific purpose behind this question, but hopefully of interest to some out there and not inappropriate for the forum.

I was wondering whether any plugins exist to allow an audio file to be opened as an image, and then subsequently re-saved as an audio file. The purpose would be to allow visual editing of sound files.

I got the idea from the Audacity software, which which can display sound information as a ‘frequency map’. In this display, the x-axis is time, the y-axis is sound frequency, and amplitude (i assume) is represented by the pixel colour value. Picture below.
image

would be interested if anyone knows of any capacity to open sound files in a visual medium and re-save them?

best
Isaac


#2

Good day,

what you showing is called a spectrogram.
Because the phase is not displayed in this kind of signal representation, you can’t reconstruct the sound from it.

Regards

Herbie


#3

Have you tried the program called “Virtual ANS” ?

Within some constraints, it can do some spectrogram processing and play it back as sound. You can also load images and use them as spectrograms and a few other very interesting things.


#4

Thanks for the info Herbie. I don’t know much about how phase is encoded, is it possible it could be represented in a separate slice/image which might then be recombined when the audio file is written?

Best,
Isaac


#5

Thanks for the suggestion gabriel, I will check that out.

best
Isaac


#6

Isaac,

your question can’t be easily answered.

Classically, spectrograms show the short-term Fourier power spectrum of the temporal signal on the vertical axis. Although it is possible to show the complex-valued spectrum, reconstruction becomes rather complicated because the temporal window has a considerable impact on the short-term spectra. Furthermore, you have to take care about the uncertainty relation of the Fourier transformation, i.e. of the fact that the more narrow the temporal window the lower is the frequency resolution and vice versa. Overlapping temporal windows don’t decrease the complexity of the problem.

Good luck

Herbie