I have been reading a paper on the "Single pass spectrogram inversion"

and I came across this in the Introduction part.

In many applications, the analysis and modification of the Short-Time Fourier Transform (STFT) and the Short-Time Fourier Transform Magnitude (STFTM) of speech and audiosignals are necessary. These applications include, but are not limited to audio enhancement, reverberation analysis, time and pitch modification, and noise cancellation. Phases are either lost, become meaningless as the spectral representations are manipulated, or simply do not exist for artificially constructed spectrograms. The objective, then, is to use these spectral representations to generate a real-valued signal that corresponds as closely as possible to the original spectrograms.

I do not understand why phases do not exist for artificially constructed spectrograms. A spectrogram simply represents what frequencies exist for a certain duration of time, right? So technically, we could just have zero phase for all the frequencies involved and it would be a realizable signal.

Any help is appreciated

  • $\begingroup$ That literally just means someone created a picture or a 2D function, said "this is a spectrogram", but didn't specify what the phase at each point was. $\endgroup$ – Marcus Müller Jun 4 '19 at 5:39
  • $\begingroup$ But even if they did not specify it, why can't an arbitrary phase be assigned to the spectrogram? $\endgroup$ – Paddy Jun 4 '19 at 5:41
  • $\begingroup$ it can. That's not what the sentence is about. Prior to assignment, it doesn't exist. $\endgroup$ – Marcus Müller Jun 4 '19 at 5:46
  • $\begingroup$ Ah, got it. I just misunderstood what it meant then $\endgroup$ – Paddy Jun 4 '19 at 5:58

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.