meandering

[dsrtao]

I wonder what the bandwidht of the sense of touch is.

Back-of-the-envelope estimate:

Update rate of 100Hz is probably adequate.
Sampling depth: 8 bits of pressure, 8 bits of temperature, 8 bits of pain.
Resolution: the highest is on the fingertips, tongue, face. Let's say that 600dpi will work on those regions, and most of the body is content with 50dpi. Let's say, 50 square inches of high-res, and 3000 of low-res.

rate * depth * resolution = 6x10^12 bits/second. That looks like a good upper bound for full-body tactile sensory transmission. I bet a simple redundancy compressor could take a big chunk of that, and a perceptual coder is probably reasonably easy.

Comments

[goldsquare.livejournal.com]

A web-page collection on Haptics papers.

[autopope.livejournal.com]

Going from memory of too many physiology lectures ...

Update rate of sensory nerves is around 20Hz, at best. You can probably make do with 10Hz. 50Hz is overkill.

Sampling depth: 8 bits is overkill, again -- I suspect you can probably do the whole lot in 16 bits (excessive pressure or temperature as a multiplier for pain).

Resolution: fingertips, tongue, face, you're talking maybe 20dpi. Rest of body -- for example, your arms or legs -- are down to about 3-4dpi. You can test this yourself: take two pins and stick them in your arm a short distance apart. See how far apart they have to be before you can sense them as separate inputs.

On the other hand, you missed out completely on internal proprioception -- sense of pain/pressure for internal organs. I guess we can't easily stimulate that without some kind of direct neural interface, but it's fun to speculate about. I suspect the entire bandwidth of the abdominal cavity is something less than 1Kb/sec, though.

[dsrtao]

Which roundly comes out to a T1, uncompressed.

The transducers, are a problem everywhere. And who wants a jack in their head without a much more reliable security system than any recently implemented?

The point of this exercise, incidentally, was to debunk the meme that VR users would drop haptics in order to save bandwidth. In any sensible situation, a small degradation in visuals will more than cover the entire bandwidth cost of haptics.

[metahacker.livejournal.com]

For reference, the military is currently testing a haptic output with a bandwidth of something like 16 baud. (8 binary outputs, updates less than 2/sec.) While you might be able to sense things more frequently than that, it takes a great deal of attention to notice and attend to transitory signals. Really, "vibrate" or "not" in each location is a good goal; perhaps three or four levels of intensity ("twitch", "vibrate", "strong vibrate", "throb"?)

[alexx-kay.livejournal.com]

I recently saw an article in New Scientist which claimed "a realistic touch interface must ideally be able to change 500 times per second or more." No idea whether this claim is backed by real data or not, but thought the article might be of interest.