0

I am working on a radio and we spend so much time to archive and upload the shows to our website, so we are searching for an "automatic" solution for that. What I came up with is to introduce a certain frequency to the starting and ending spot of every show (probably about 21khz, I am aware that FM transmitters work from 50 to 15khz, but the recorded files will be straight from the mixer, not from the radio transmission).

After that, I would duplicate the file, filter it out with a bandpass filter so i have the same file length but now only with the 21khz points of sound. This could be done with a python script or with Nyquist, although I am not sure if this can be automated through command prompt with nyquist.

Next, and here is my problem, I would need to export a label (audacity) file or a .XMCD, .CDDB or .CUE file with the exact time those frequencies occurred. at the end I would use mp3splt to split the first "original" file with the label or cue or CDDB or XMCD file I created.

Till now I managed to do this with audacity, thanks to that advice I found. I also read about frequency detection in python and about filtering a recording in python and also found this interesting way in audacity forum for importing timestamps using python

What do you think would be the best way to continue with that problem? Would I be able to do it with python? Maybe with something similar to silence detection?

The label text file that audacity exports is in the form of:

time 1 {tab} time 2 {tab} "a character" 

time 3 {tab} time 4 {tab} "another character"

or 

time1 {tab} "a character" 

time2 {tab} "another character"

thanks in advance for any help

Community
  • 1
  • 1
user3426993
  • 1
  • 1
  • The solution you are searching for is exciting. But after reading your question completely, I feel that you are going to implement a complex solution. Too complex I think. Keep it simple as stupid (kiss). You need a dedicated communication channel to declare start and stop information. You want to use an inaudible frequency (21kHz) to carry this information. Why not using another channel, like a dedicated file of timestamp (Unix timestamp) for example, that would be shared between emitting and archiving process? – RawBean Mar 17 '14 at 12:17
  • I am not sure it is that complicated. You can use a short inaudible sync symbol (such as LFM chirp) and produce the time record using cross-correlation with the known signal. The unique sync symbol will ensure no false alarms. – ThP Mar 19 '14 at 20:50

0 Answers0