Skip to main content

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: Human cochlea to brain bandwidth is 10-20kbps (Read 3540 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Human cochlea to brain bandwidth is 10-20kbps

Hi all,

I once read an article saying that the auditory nerve connecting the cochlea to the brain has only enough bandwidth for 10-20kbps, (somewhere thereabouts) providing the theoretical basis for the search for an lossy encoding algorithm that can output completely transparent audio at drastically lower bitrates than CD, theoretically as low as the bandwidth of the auditory nerve itself. Since then I've quoted this every now and then when people doubted the premise behind lossy encoding, but haven't been able to find the original reference. Does anyone know where I can find it?

Human cochlea to brain bandwidth is 10-20kbps

Reply #1
If it exists you should be able to find it using this search engine.
Let's suppose that rain washes out a picnic. Who is feeling negative? The rain? Or YOU? What's causing the negative feeling? The rain or your reaction? - Anthony De Mello

Human cochlea to brain bandwidth is 10-20kbps

Reply #2
Hmm. Dunno how they would calculate something like this.

I suspect you may have misread "10hz-20khz" somewhere!

 

Human cochlea to brain bandwidth is 10-20kbps

Reply #3
hmm... find a gullible volunteer and wire an ADC to the other end of their auditory nerve?

hmm.  isn't it a huge array of nerves rather than a single one?  that would mean 10-20kbps per "hair", in which case there's no hope for lossy codecs going that low transparently.

Human cochlea to brain bandwidth is 10-20kbps

Reply #4
Quote
Hi all,

I once read an article saying that the auditory nerve connecting the cochlea to the brain has only enough bandwidth for 10-20kbps, (somewhere thereabouts) providing the theoretical basis for the search for an lossy encoding algorithm that can output completely transparent audio at drastically lower bitrates than CD, theoretically as low as the bandwidth of the auditory nerve itself. Since then I've quoted this every now and then when people doubted the premise behind lossy encoding, but haven't been able to find the original reference. Does anyone know where I can find it?

There's a discussion of "the data rate problem" in the following paper...

Auditory Models as Preprocessors for Speech Recognition
R D Patterson
J Holdsworth
M Allerhand
in
Speech Research 10 - The Auditory Processing of Speech - From Sounds to Words
published by Mouton de Gruyter, Berlin NY 1992

...but it doesn't state what you say - it just compares the data rate at the outputs of various auditory models. I had a feeling that Patterson had another paper which directly compared this to the actual human auditory system, but I may have been wrong.


However, the idea that the data rate on the auditory nerve is only 10-20kbps, therefore that could be the target datarate of psychoacoustic based coding may be totally misleading...

For one thing, we can more our heads while listening. This changes the sound pressure wave that actually impacts our ears. At any moment, we sample the sound field at one point in space, but just by turning our head slightly, we receive completely difference signals. So, that's one piece of "redundancy" that we can't actually remove.

For another, we can adjust our attention. It's likely this is a higher brain activity, but it's just possible (it's beyond my knowledge - someone may be able to clarify) that feedback from the brain can effect what the ear transduces. The fact that there's more than one feedback loop in operation is widely accepted, but does the action of that feedback depend in any way on our conscious listening (rather than hearing) process? If so, then the 20kbps that make it to the brain are partly chosen by the brain - and another 20kbps could easily have been chosen instead. So, both the person talking, and the background noise, need to arrive intact at the ear, because we could choose to listen to either, while that 20kbps may only contain one. (This is no more than an idea - and it could well be wrong)

Finally, it's quite likely that the lossy transductions in the ear+brain will only give a particular result with a particular input. It seems, in theory, that if you know that a certain 20kbps is going to occur on the auditory nerve, then you should be able to send the right signals to loudspeakers to cause this, and these signals should be generate from a 20kbps data stream. However, that's assuming that we can reverse the ear+brain process perfectly, which is unlikely.

Imagine, say, a VHS video tape. Recording something onto that is certainly a lossy process! Imagine the VHS tape is our 20kbps representation of the signal. If you start with a VHS tape, and try to copy it onto a second VHS video tape, the result looks even worse. Or imagine a 128kbps mp3 file, decoded. It's very hard to re-encode it to a 128kbps mp3 file without further loss of quality. So, if you want a "decent" VHS copy, you start with a better quality master. If you want a "decent" 128kbps mp3, you start with a better quality master. If you want the brain to create a good 20kbps auditory nerve signal, you have to feed it something better. Unless you understand the entire process perfectly, and can reverse it.


However, the limited datarate on the auditory nerve is still a justification for psychoacoustic based coding, because it does place limits on how good a signal needs to be.

The fundamental limits of the transduction process within the ear itself are actually the basis for psychoacoustic based coding, and so these form a much better justification in an argument. A real physical membrane moves in the ear, and hair cells are triggered by this movement. If two different signals cause exactly the same movement, or if the difference is way below the noise floor of the hair cells, then you can't hear the difference. It's as simple (but, in reality, as complicated!) as that.

Cheers,
David.

Human cochlea to brain bandwidth is 10-20kbps

Reply #5
Thanks dave!

Human cochlea to brain bandwidth is 10-20kbps

Reply #6
The ear's response has nothing to do with the speed of transmission from ear to brain.  Sounds excite cilia in the cochlea which respond to different frequencies.  The ear transducer's response is then frequency-,  not time-based,  making the transmission speed irrelevant.
The brain is not a von-neumann computer folks. 

Cheers

Madman

Human cochlea to brain bandwidth is 10-20kbps

Reply #7
Quote
The ear transducer's response is then frequency-, not time-based


And does that mean no data bandwidth is required to send the auditory information in frequency-based format?