What? News of a possible 3D sensor? Sort of. Here’s the deal.
20161219/https://youtu.be/0QNiZfSsPc0
20161219/https://atap.google.com/soli/
The video touts the radar as “very high positional accuracy,” which is more vague than deserved, so here is an accurate number computed from the specification on the Project Soli website. 60 GHz electromagnetic radiation has a 5 mm wavelength, so being optimistic about sub-wavelength detail recovery algorithms, “very high positional accuracy” is measuring details within 1 mm of accuracy.
Also note that the portion of the video displaying the reflected waves off of the hand to the sensor was drawn to spatial scale, given the wavelength of the radar signal. Time scale, of course, was greatly exaggerated.
Before you go too far “imagining the possibilities,” you still have to remember the limits.
This is in fact “microwave” radiation.
20161219/https://en.wikipedia.org/wiki/Microwave
Oh, note a limitation with microwave radiation is atmospheric transmissiveness to it. The particular frequency selected for the Google sensor is strongly absorbed by the atmosphere, but other nearby low frequencies are very transmissive through the atmosphere. Higher frequencies, not so much.
Right after microwave radiation is far infrared. Then again, wait, actually there is not a formal distinction and there is overlap between different terms used in different fields of study.
Oh wait, now I am told the sensor is capable of detecting with sub-millimeter accuracy. Yes, but there’s a difference between “can” do and “do well.” For scientific reasons, there are obvious limits as to how far this statement can be stretched.
20161219/https://en.wikipedia.org/wiki/Google_ATAP#Project_Soli
Interesting, the Wikipedia article briefly mentiosn all of the related techniques in a single sentence. Wow, I’m impressed actually. You know, it’t too bad that the other article is not as well written.
Yes, yes, yes, just like in all the literature you’ve read about semiconductor manufacturing. The minimum feature size is a scalar of the wavelength of light used.
-
Hey! That’s it! Since you always use these 3D sensors by specifying a minimum feature size and a maximum distance, that works great for digital computer arithmetic! No need to worry about arbitrary precision integers when you know the scope of your scanning equipment. But then when you combine it all together… yeah, that’s when you’ll need to consider that.
-
And definitely. By the time I get around to my prototype, all of this will be old hat, not very interesting, and very well known and understood by the future community.
Oh yeah, but definitely note what it says on the website.
Unlike traditional radar sensors, Soli does not require large bandwidth and high spatial resolution; in fact, Soli’s spatial resolution is coarser than the scale of most fine finger gestures. Instead, our fundamental sensing principles rely on motion resolution by extracting subtle changes in the received signal over time.
A big difference between 3D scanning versus 2D photography is that 3D scanning tends to be distance and resolution limited based off of the parameters of the scanner, whereas visible light’s resolution and distance often times depends only on the optics used.
- Now, time for a showcasing of consumer electronics that implement each of these techniques, or N/A if I could not find one. Seriously, this is something that just couldn’t have been done a couple of years ago.
** Yes, although the technologies have been around for quite some time, they have long been locked up in extremely expensive industrial equipment. Some of them still are, but many of them are fast approaching the mass market.
- Yeah, of course you can make a subject-level categorization, but often times it makes more sense to rank the available sensors by their costs. Because, in the real world, you can go and get low-cost equipment right away, but expensive things, you’re going to have to think about that for a while. Also, once cost is pinned, the next questions to think about are quality and speed.
And, I have an interesting thing to say of my own.
A picture is worth 10,000 words but a gesture is only worth 50 or less words. Sure, that makes a gesture recognition device 200 times easier to build (READ: 200 times cheaper) than a generic radar imaging device, but it also means that the devices are distinctively limited in their usefulness.