What sort of inaudible audio signal survives compression?
September 14, 2008 7:25 PM Subscribe
I recently became an Arbitron ratings person (which I'm not allowed to tell anyone, hence my anonymity). I'm supposed to carry a portable device that they say picks up "sound energy" signals embedded by participating radio and TV stations so they know what I'm listening to, even if the audio is compressed. What the heck is "sound energy"?
Here's what really has me confused: This "sound energy" (their term) is inaudible to humans. But according to Arbitron, their gadget even pick up the signal on podcasts or streaming internet radio. But doesn't audio compression work by (among other things) stripping out all the inaudible portions of the audio? So what's left for the device to pick up?
And what kind of "sound energy" can be picked up by the device whether I'm listening on high quality speakers, or a tinny transistor radio? And the device doesn't even need to be near the speaker. They say that if my ears can hear the radio, even from another room, so can the device.
So how the heck does this thing work? Is Arbitron overstating the efficiency of their device, or is there some magic technology at work?
Here's what really has me confused: This "sound energy" (their term) is inaudible to humans. But according to Arbitron, their gadget even pick up the signal on podcasts or streaming internet radio. But doesn't audio compression work by (among other things) stripping out all the inaudible portions of the audio? So what's left for the device to pick up?
And what kind of "sound energy" can be picked up by the device whether I'm listening on high quality speakers, or a tinny transistor radio? And the device doesn't even need to be near the speaker. They say that if my ears can hear the radio, even from another room, so can the device.
So how the heck does this thing work? Is Arbitron overstating the efficiency of their device, or is there some magic technology at work?
It sounds like someone is using the word "sound energy" to mean Audio Watermarking of some kind, but even that sounds very, very overstated.
posted by majick at 7:59 PM on September 14, 2008
posted by majick at 7:59 PM on September 14, 2008
You could always look on Arbitron's site and find out.
posted by Lazlo at 8:10 PM on September 14, 2008
posted by Lazlo at 8:10 PM on September 14, 2008
Sounds kind of creepy, actually...but Lazlo beat me to the link. I think thats your best resource unless you talk to an audio engineer.
posted by saxamo at 8:16 PM on September 14, 2008
posted by saxamo at 8:16 PM on September 14, 2008
This looks like the Arbitron Portable People Meter?
They're being overblown by saying "sound energy." They've just slightly modified the sound being broadcast.
But doesn't audio compression work by (among other things) stripping out all the inaudible portions of the audio? So what's left for the device to pick up?
Lossy audio compression does try to do this, but I immediately see two ways around this, and it's not clear to me which they are using. The first is that they could design an imperceptible signal which passes through the widely implemented lossy compressions. The second is that their encoder could be designed to come after lossy compression.
More generally, issues of cheap speakers, distance from speakers, other rooms, to some degree compression, etc. are addressed by the Shannon capacity of the channel. Roughly, while the specific formula shown there isn't quite what's going on in this case, the bits/second rate of a communications channel increases with bandwidth and signal strength, and decreases with noise strength. These things are constrained in Arbitron's case, where the channel is that between the Arbitron Encoder and the Arbitron Portable People Meter. We can consider the actual music signal (as well as the actual radio and other noise) as "noise," the strength of the Arbitron signal is constrained by inaudibility, and the bandwidth is constrained in numerous ways primarily based on the 0-20 kHz response of the human ear, though I'll bet Arbitron uses even less for various reasons. But even with the numbers unfavorably constrained, there is a channel capacity, and Arbitron needs very few bits/second - for their purposes it doesn't matter if they're measuring speed in seconds per byte instead of bytes per second.
Also at issue is even if their technology works only poorly, they're hardly going to tell everyone.
posted by TheOnlyCoolTim at 8:22 PM on September 14, 2008
They're being overblown by saying "sound energy." They've just slightly modified the sound being broadcast.
But doesn't audio compression work by (among other things) stripping out all the inaudible portions of the audio? So what's left for the device to pick up?
Lossy audio compression does try to do this, but I immediately see two ways around this, and it's not clear to me which they are using. The first is that they could design an imperceptible signal which passes through the widely implemented lossy compressions. The second is that their encoder could be designed to come after lossy compression.
More generally, issues of cheap speakers, distance from speakers, other rooms, to some degree compression, etc. are addressed by the Shannon capacity of the channel. Roughly, while the specific formula shown there isn't quite what's going on in this case, the bits/second rate of a communications channel increases with bandwidth and signal strength, and decreases with noise strength. These things are constrained in Arbitron's case, where the channel is that between the Arbitron Encoder and the Arbitron Portable People Meter. We can consider the actual music signal (as well as the actual radio and other noise) as "noise," the strength of the Arbitron signal is constrained by inaudibility, and the bandwidth is constrained in numerous ways primarily based on the 0-20 kHz response of the human ear, though I'll bet Arbitron uses even less for various reasons. But even with the numbers unfavorably constrained, there is a channel capacity, and Arbitron needs very few bits/second - for their purposes it doesn't matter if they're measuring speed in seconds per byte instead of bytes per second.
Also at issue is even if their technology works only poorly, they're hardly going to tell everyone.
posted by TheOnlyCoolTim at 8:22 PM on September 14, 2008
According to wikipedia this is done via psychoacoustic masking.
The idea here is that your human ear>brain can be fooled into hearing a vocalist but a few ms before he sang there was 5000herz tone that your mammalian brain cannot hear, even though it is clearly being played and within the sound range of your ear.
A lot more on auditory masking here. Masked sounds should survive typical music compression.
posted by damn dirty ape at 8:29 PM on September 14, 2008
The idea here is that your human ear>brain can be fooled into hearing a vocalist but a few ms before he sang there was 5000herz tone that your mammalian brain cannot hear, even though it is clearly being played and within the sound range of your ear.
A lot more on auditory masking here. Masked sounds should survive typical music compression.
posted by damn dirty ape at 8:29 PM on September 14, 2008
NYT article:
So Kolessar began to work on psychoacoustic masking, which places a sinal just beneath the frequency of whatever is being transmitted. As Kolessar and his team worked through years of frustration, they discovered that the masked code's frequency could not be too low (where it would run into technical problems) or too high (where it would bother dogs and cats). Nor could it even modestly compromise the audio quality of a show or a song. ''We used 'Achy Breaky Heart' for a while for our tests, but then I just couldn't take it anymore,'' Kolessar said. So he switched to ''Don't Give Up,'' the Peter Gabriel and Kate Bush duet from Gabriel's album ''So.'' The song is an intricate work of sonic architecture (of some 26 separate tracks, according to Kolessar), which made it a challenge for the engineers. Kolessar led me into Arbitron's sound studio to listen to what his team came up with after a decade. He cranked a recording of the Gabriel song through a pair of $20,000 speakers and switched back and forth between coded and uncoded versions. ''You can't tell the difference,'' he said, more as a statement than a question. I agreed.
posted by damn dirty ape at 8:33 PM on September 14, 2008
So Kolessar began to work on psychoacoustic masking, which places a sinal just beneath the frequency of whatever is being transmitted. As Kolessar and his team worked through years of frustration, they discovered that the masked code's frequency could not be too low (where it would run into technical problems) or too high (where it would bother dogs and cats). Nor could it even modestly compromise the audio quality of a show or a song. ''We used 'Achy Breaky Heart' for a while for our tests, but then I just couldn't take it anymore,'' Kolessar said. So he switched to ''Don't Give Up,'' the Peter Gabriel and Kate Bush duet from Gabriel's album ''So.'' The song is an intricate work of sonic architecture (of some 26 separate tracks, according to Kolessar), which made it a challenge for the engineers. Kolessar led me into Arbitron's sound studio to listen to what his team came up with after a decade. He cranked a recording of the Gabriel song through a pair of $20,000 speakers and switched back and forth between coded and uncoded versions. ''You can't tell the difference,'' he said, more as a statement than a question. I agreed.
posted by damn dirty ape at 8:33 PM on September 14, 2008
Also this:
Arbitron has begun to ask radio and television stations around the country to run their broadcasts through a patented Arbitron encoding device; at the moment, almost all of the radio and television stations that a listener can tune into in the Houston metro area -- including over-the-air, cable and satellite TV (though not satellite radio) -- are coded for the P.P.M. trial. The stations are not being paid for this; instead, Arbitron has convinced them, through literal door-to-door salesmanship, that encoded broadcasts will enable Arbitron to measure their audiences better and thereby ultimately boost their advertising sales.
So this device is somewhere on the airchain of the broadcast. No need to re-rip all that music.
posted by damn dirty ape at 8:43 PM on September 14, 2008
Arbitron has begun to ask radio and television stations around the country to run their broadcasts through a patented Arbitron encoding device; at the moment, almost all of the radio and television stations that a listener can tune into in the Houston metro area -- including over-the-air, cable and satellite TV (though not satellite radio) -- are coded for the P.P.M. trial. The stations are not being paid for this; instead, Arbitron has convinced them, through literal door-to-door salesmanship, that encoded broadcasts will enable Arbitron to measure their audiences better and thereby ultimately boost their advertising sales.
So this device is somewhere on the airchain of the broadcast. No need to re-rip all that music.
posted by damn dirty ape at 8:43 PM on September 14, 2008
Whichever instruments you're using to receive these transmissions, they're being transmitted at a frequency which isn't perceived by human hearing, but who knows what kind of impact they have otherwise. That these frequencies would be mixed together with AM and FM is a tad freaky. According to wiki: *An encoder inserts the tones (via psychoacoustic masking) into a station's or network's airchain.* Psychoacoustic masking seems to be where you need to researching this further.
posted by watercarrier at 2:19 AM on September 15, 2008
posted by watercarrier at 2:19 AM on September 15, 2008
It's never a bad idea when a corporation or business firm is involved to consider fraud. Maybe, and I'm just sayin', maybe Arbitron has a contract with the broadcasters to produce specific results, and they have built a device like the Scientologist's brain analyzers which are guaranteed to produce the results they want.
posted by vilcxjo_BLANKA at 6:41 AM on September 15, 2008
posted by vilcxjo_BLANKA at 6:41 AM on September 15, 2008
This thread is closed to new comments.
posted by mpls2 at 7:33 PM on September 14, 2008