Just starting research in audio watermarking and have read many papers regarding audio watermarking but I am still a little confused on what exactly gets embedded into the signal. For example, one paper I read embedded an image or written signature into the audio.

My question is what typically gets embedded in the signal using audio watermarking? Is it just random data, a picture, someone's name? Thank you!

  • $\begingroup$ It's whatever you want it to be $\endgroup$ – endolith Jul 2 '13 at 18:52

Research papers or textbooks won't delve into the specifics - they just assume you have a stream of binary data that needs to be reliably and undetectably embedded into the audio file.

What this bitstream is actually going to be is application dependent. It could be:

  • Digital rights management applications: A short serial number identifying the individual transaction in which the audio file has been sold. Given the small size (a hundred bits) of this information, it is typically repeated over and over the file to maximize the probability of recovering it.
  • Metadata management: Metadata such as artist name, song title, or even lyrics.
  • Proper steganography: Any secret data you want to share with the recipient of the audio file - without an eavesdroppers suspecting that you are exchanging anything other than music. This can be a text-file, an image - any binary data.
  • Informed source separation: Any information about the individual sources the audio file is made of, that would subsequently help in recovering these individual sources.
  • Second-screen interactive TV applications: These application allow a device such as a smartphone or a toy to display secondary information synchronized with a TV show. To allow the synchronization, the audio signal of the show contains a regular watermark containing a show identifier and time stamp.
| improve this answer | |
  • $\begingroup$ Thank you for a great answer. That explains a lot. I am interested in the metadata management. How do you turn, for example, an artist's name into data so that it can be embedded into the audio file? $\endgroup$ – Math244 Jul 2 '13 at 15:10
  • $\begingroup$ I am not aware of industry standards for this application, but if you want to implement your own solution, it could be as simple as ASCII-encoding the data using delimiters between each field... $\endgroup$ – pichenettes Jul 2 '13 at 17:54
  • $\begingroup$ Can you use ASCII to encode the data when the audio signal is in double precision? Wavread in matlab outputs the data in double precision. $\endgroup$ – Math244 Jul 3 '13 at 4:02
  • $\begingroup$ I don't understand your problem - the format in which the audio data is and the sequence of bits you 'hide' inside it are two different things. Since this is a programming/implementation question, it is out of topic on this site anyway... $\endgroup$ – pichenettes Jul 3 '13 at 9:10
  • $\begingroup$ Okay well thank you very much for all the information. It's been a great help. $\endgroup$ – Math244 Jul 3 '13 at 12:50

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.