I very simply have a .wav audio file, but there is a note with it that says "This audio has been sped up by 10x".

I am wondering, what does it mean exactly from a signal processing perspective for an audio signal to be 'sped up' by some factor, and secondly, how can I undo this? Is it possible?

  • $\begingroup$ If it wasn't tempo-preserving algorithm then simply divide your sampling frequency by 10 and you are done. On the other hand, when you want to make the Alvin and the Chipmunks sound, then record a song and save it with twice or more higher sampling frequency ;) $\endgroup$
    – jojek
    Jun 4 '14 at 13:51

It could mean:

  • That the metadata in the .wav file wrongly indicates that the signal was recorded at a sample rate 10x higher than the true value (for example it is advertised as having been recorded at 80kHz while it has been recorded at 8kHz). Simply fixing the metadata undoes this.
  • That the signal has been downsampled by a factor of 10, but is still tagged as having been recorded at the original sample rate. This can be undone by dividing by 10 the playback sample rate (by tweaking the metadata); or by upsampling the signal by a factor of 10 without changing the playback sample rate. Note that in any case, 90% of the bandwidth of the original signal (the high frequency content) has been lost and cannot be restored (or only very approximately through spectral band replication).
  • That a time-stretching (time-scale modification) algorithm has been applied to it. This is a lossy process, and cannot be undone without severe loss at such a high stretch factor. The "antidote" is simply to use any time-stretching algorithm with a stretch factor of 10x.
  • $\begingroup$ Hmm, well, I guess I just dont know. Here is the file and its note. I dont know which possibility it is... $\endgroup$ Jun 4 '14 at 14:07
  • $\begingroup$ @TheGrapeBeyond: I am almost 100% sure that they simply saved file with 10x higher sampling frequency. That's because whales are producing very low frequency sounds, thus if we want to hear it, we must shift the pitch up. Easiest way --> cheat on sampling frequency. If you want to retrieve original signal, do the reverse, i.e. in Audacity. Aaaaa, what the heck - here is your original file. $\endgroup$
    – jojek
    Jun 4 '14 at 14:22
  • $\begingroup$ I've noticed the file has a discontinuity every 400th sample, which might indicate that time-stretching has been applied simply by chopping the signal into 400 samples-long blocks, and keeping only every 10th block. This is going to be a tough case because the first thing you'll have to do is get rid of those horrible clicks. $\endgroup$ Jun 4 '14 at 14:27
  • $\begingroup$ @jojek Thanks jojek! I will look into this file... $\endgroup$ Jun 4 '14 at 14:42
  • $\begingroup$ @pichenettes Hmm, thats interesting... I guess I would not know..., what about the simply-changed-fs hypothesis? $\endgroup$ Jun 4 '14 at 14:43

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.