Audio Watermarking Example - Can You Tell the Difference? (Listening Test)
You may have heard of the different audio watermarking techniques by now. Audio watermarks allow for audio from musicians, broadcasters and others to be tracked, measured and can be used to engage audiences.
Even if you understand the theory behind the audio watermark, you might be wondering what the practical implications are.
In this post, we take you through the differences between the popular technologies and then let you hear the difference for yourself with some audio watermarking examples.
We recommend finding a quiet spot and getting your headphones on to really hear the difference!
The Differences in Audibility For Various Watermarking Types
We have already written an in-depth guide about whether you can hear audio watermarks where we discuss the three technologies below in more detail.
Here’s a brief recap of the technologies before we show you some examples you can listen to.
Ultrasound watermarking gained popularity for being generally inaudible. The metadata can be hidden outside of the human-perceivable spectrum using this technology.
While this solves user experience issues for listeners, this method isn’t very robust and is removed by compression algorithms rendering it unusable for broadcasters who want to broadcast over platforms like YouTube or Netflix.
To make matters worse, this type of audio watermarking doesn’t survive over traditional broadcasting platforms either, such as satellite and cable.
- Data is hidden beyond the 16kHz “ultrasound” range making it inaudible
- Ultrasound watermarks are stripped out by compression algorithms, including the likes of YouTube, Vimeo, BBC and Netflix streaming services
Spread Spectrum Watermarking
Spread Spectrum came after ultrasound technology to solve this lack of robustness.
Probably the most famous example of spread spectrum watermarking use is by Universal Music Group.
While that name alone might not mean much, they are responsible for distributing around 25% of most music you listen to on radio, iTunes, Spotify and other music streaming services.
This type of watermark is spread across the entirety of the audio (hence, the name “spread spectrum”) and is concentrated in the 1 khz and 3.5 khz frequency range which makes removal of the watermark more difficult.
But it also means that some sensitive listeners can hear this watermark and complain of a “hum” while listening to their favourite tracks or broadcasts.
- More robust than ultrasound watermarking and more difficult to remove, even with compression
- Sensitive users complain that the watermark is audible and ruins their listening experience
- The watermark itself is sensitive to the Doppler Effect. This is the pitch shift that occurs if the user is moving while listening to audio. The data from the watermark may fail to be extracted if the user isn’t completely still which can make it impractical
Echo modulation watermarking takes advantage of the evolution of the human auditory system evolving to filter out short echoes.
The data contained in the watermark is artificially added to and hidden in the natural echoes of the audio being watermarked.
This offers the best of both being practically inaudible to the human ear, as well as robustness.
Intrasonics offer patented echo-modulation audio watermarking technology developed in-house to solve the problems that other audio watermarking technologies suffer from.
- Short, inaudible echoes hidden within the natural echoes of the watermarked audio.
- More data can be stored using multiple, non-overlapping echoes.
- More robust - less susceptible to the Doppler effect which means moving devices can still decode the watermarks.
- May be audible in some situations such as a classical piece that uses a single instrument.
Audio Watermarked Samples - Can You Hear Them?
Let’s get to the fun stuff and see if you can actually hear the different watermarks.
We recommend putting your headphones on in a quiet environment for the best results.
Echo Modulation Watermarking Examples
We’ll start with the technology which we’re known for here at Intrasonics; echo modulation.
Here are some echo modulation watermarking examples on music tracks of different genres.
Which tracks or genres can you hear the audio watermark on?
Dubstep - Original (no watermark)
Dubstep - Watermarked
High Energy - Original (no watermark)
High Energy - Watermarked
Epic - Original (no watermark)
Epic - Watermarked
Happy Rock - Original (no watermark)
Happy Rock - Watermarked
Inspirational - Original (no watermark)
Inspirational - Watermarked
Pop/Dance - Original (no watermark)
Pop/Dance - Watermarked
Slow/Sad - Original (no watermark)
Slow/Sad - Watermarked
Happy/Upbeat - Original (no watermark)
Happy/Upbeat - Watermarked
Which genres of music did you recognise the audio watermarks in, if any?
Echo-modulation watermarks are almost inaudible in audio where we experience many echoes - for example, in a live sports match.
Genre’s where they might be more audible are ones where our ears are not as accustomed to echoes. For example, single-instrument classical pieces of music.
The watermarked tracks are encoded with CH8 and codeword 32800. This means the tracks can be decoded using our decoder which you can access in our developer portal HERE to see how decoding the watermark works.
Spread Spectrum Watermarking Examples
We found the best spread spectrum watermarking samples on Matt Montag’s website.
Matt is a software engineer at YouTube and Google. You can take his listening test here and compare over 15 audio tracks with spread spectrum watermarks.
If you want to skip to the spoilers, you can see how other people were able to perceive the watermarks for each track here.
Electronic music with strong transients was the music genre where spread spectrum watermarks were least able to be detected by the human ear, whereas most people heard them when listening to classical, orchestral and piano music.
Things to Consider and the Trade-Off’s
Of course, it’s important to know the theory but in the real world, what are the things to consider when choosing which watermarking technology you go for?
We feel that the trade-off is between data rate, robustness and audibility of the watermark, as visualised below.
If a watermark is designed to be completely inaudible to the human ear, you are likely to sacrifice data rate and robustness.
In a perfect world, watermarks would be completely silent but if the strength of a watermark is too low, then very little data can be sent. This also makes it less robust and easier to be lost in compression.
Maximum robustness means more data is needed for error correction (this ensures that even partially received watermarks can be correctly extracted), but this doesn’t lend itself well to a completely inaudible audio watermark.
Ultimately, it’s clear from our diagram that there’s no such thing as the perfect audio watermark.
At Intrasonics, while we feel that our echo modulation technology gives you the best of the three worlds straight out of the box, we also design our watermarks with the configurations that are most important for your specific use-case.
We call these configurations “schemes” and we’ll help configure a watermark for the trade-off that best fits your need.
< Previous Page