The development of our watermarking technology began at the end of the 90s. The aim of the development was an imperceptible yet robust watermark. The process was to be capable of being applied to different audio formats so that the same software can always be used for the analysis of an audio signal for a watermark without prior knowledge of the embedding process used being required.
For embedding the watermark, a so-called psycho-acoustic model is used. This determines the amount of disturbance which can be introduced into the particular part of the audio signal without the disturbance being audible. Thus the approach is very similar to an audio encoding procedure like MP3 only there the disturbance results from data quantity reduction but with our watermarking process, the disturbance carries the desired additional information.
After the introduction of the additional information, the audio signal is transferred again in the original format. The watermark is completely compatible with the audio formats supported so the result can be reproduced with the same playback devices as the original. Thus it is also clear that the watermark does not represent hard copy protection. If however, copies are found on the Internet, the original recipient, that is the person who first circulated the track, can be indicated with the help of the embedded information.
The watermark is embedded in the audio signal in data containers. The data container format allows for the transmission of 48 bit usage information in 5 second intervals (at 44.1kHz sampling rate). 48 bits may not sound like a lot but with that, 281,474,976,710,656 distinguishable copies of a track can be made. So, every person in the world could receive over 35,000 copies of the same track.
The information which is to be embedded can be determined for every track or for every track copy. Our customers decide for themselves whether this information is a customer number, a transaction number or maybe a combination of both. The information is then re-embedded until the end of the file is reached. So, in a 3 minute long piece of music, the information is contained 36 times in the left-hand as well as in the right-hand channel.
Every customer is assigned their own watermark key. Consequently, each customer is only able to extract information which they themselves embedded. Conversely, accessing other customers’ embedded information is not possible. Hence, the wrong assignment to a user of a watermark belonging to someone else is precluded.
Watermark embedding and also reading takes a fraction of the playtime of the music piece. The embedding period can be further reduced if the watermark is not introduced over the whole piece but only in sections. This option is especially interesting in the production of audio books with a longer play time.