Tiny Radar Chip Revolutionizing Gesture Recognition: Google ATAP’s Project Soli | Hacker News Reader

Tiny Radar Chip Revolutionizing Gesture Recognition: Google ATAP’s Project Soli

allaboutcircuits.com

112 pointselijahparker34 comments

Small radar ICs will be useful for other things.

Radar gives you range rate directly. Getting range rate from vision is hard and takes several frame times at least. If you have position and range rate and dump it into a machine learning system, gesture recognition should be straightforward.

It's not clear how useful this will be for GUIs in general. It could be useful for interacting with large screen displays, where you may be too far back for touch and a mouse is a pain. A presentation version could be useful.

(Presentations from computers still look lame. The audience usually sees the presenter's desktop while they futz with the computer, then enter PowerPoint.)

> Presentations from computers still look lame. The audience usually sees the presenter's desktop while they futz with the computer, then enter PowerPoint.

1. Unplug VGA/HDMI cable

2. Do all of the futzing

3. Plug VGA/HDMI cable

This is a solved technical problem. That people can't be bothered is the real problem.

You have to consider how the computer reconfigures itself every time you add or remove a display device.

Except half of the time your OS does weird-ass things in step 3, displays the wrong part of the screen, wrong resolution, crashes the app, etc. It's better now than it used to be, but then again a lot of presenters don't use bleeding-edge hardware with most recent & fully updated OS.

Alternative if you present regularly:

1. Enable "extended destkop", rather than "mirror desktop".

2. Select an appropriate wallpaper, or just black.

Best i can tell, presenters use mirror so that they can see their own slides on the screen while talking.

Frankly i think a seldom talked about Android feature may be a better solution.

https://developer.android.com/about/versions/android-4.2.htm...

I think PowerPoint already does this; Keynote's had it for ages, and I've done it on Linux with Beamer a few years back. You have presenter's notes with a small slide preview on the screen in front of you, actual slides on the big screen.

These days I run my presentations in reveal.js (HTML5), so the output is trivially cross-platform and videos, 3D content etc. always work flawlessly. This also has speaker notes, implemented as two browser windows (one for each screen) with linked controls.

Any presentation tool worth using has a presenter mode where on your primary display you see things like the current slide, your notes, a timer, the upcoming slide, etc.

I though the auto industry was commoditizing this technology with minature and cheap technology.

This also happened with air bag accelerometers which sell for dimes now. They were first used to detect falling laptops and retract rotating disk arms to save disks. Then you put two or three in mobile devices figure out the 3D orientation of screens.

I always wondered how the phone knew its orientation. Then I saw this video. Wow.

https://www.youtube.com/watch?v=KZVgKu6v808

This is from a series of videos called "Engineerguy"

His videos are wonderful.

In my opinion this is awful for those of us who have to help people do things from time to time. Yet another UI that is impossible to walk somebody through over the phone.

I'm working in tech support and your post just made my hair turn white.

Yes, well, at least now you can have your computer shut down by giving it the finger.

If this is available with a few meters of range, it has potential for automatic door control. Star Trek quality door control has been done experimentally [1] but the standard sensors are too dumb. You want range, azimuth and range rate. This is no big deal in a radar; it just costs too much.

If someone is moving fast, the door opens sooner, and faster. Moving parallel to the door should not trigger it. (In Japan, automatic sliding doors on street storefronts are common, and false triggers from passers-by are a problem.) With good sensors on both sides, you can open the door further in advance if there's someone on the other side. This allows opaque automatic doors.

[1] https://www.youtube.com/watch?v=WhrLetrTSYA

60Ghz radio, 2-3 foot range.

   Soli's sensors can detect motion at a range of about two to three feet, Schwesig says, so any device you use within that range stands to benefit.

http://mashable.com/2015/05/30/google-project-soli-analysis/...

This is really exciting. It's sort of clear that there is no one killer interface, but truly robust gesture recognition, DL driven natural language interfaces, and speech recognition start to describe a nice landscape for VR and AR.

When I'm trying to create a lot of text, I'll still use a keyboard, but most of the time that's not what I want to be doing.

Yeah, most developers say they'll stick with a keyboard.

I'm still holding out for the clever person who integrates gestures into a keyboard killer. You can use gestures chords, for example. With intellisense and other intelligent completion, the future might not need a keyboard.

Here are some other gesture solutions:

https://github.com/melling/ErgonomicNotes#gesture-computing

I wonder if they have to be exclusive.

How cool would it be to have a keyboard where you can hold down a "gesture button" or just move your hands/fingers over a certain zone and have it pick up small gestures? You'd barely need to move your hands and could probably keep your wrists in the same position.

This could then do things like more natural scrolling, cursor control, obviously more complex actions.

They keyboard is great for what it does, but improving input doesn't necessarily have to mean getting rid of something that works quite well and completely reimagining it.

What has probably seen the end of the road, is the mouse.

Not yet. There's few things matching its combination of speed and precision.

It's more believable that computer mice will get tech like Soli integrated to become smarter.

So... a mouse and a motion control transceiver in one? That could be fun.

I don't think automatic completion is ever going to be a good substitute for solid text input. If your text input solution always requires it, I for one am not buying. That said, I'm hopeful for a gesture-based text input system that doesn't require any crutches.

Soli in particular seems like it could be a huge driver for this, given that the market is "Mobile devices". I'm hugely excited.

I wish there was more technical detail. I would love to know the performance characteristics of the radar, and what techniques it uses (eg. is it fully coherent, waveform repetition frequency, etc).

They've had "small" radar modules for a while, commonly used as microwave movement detectors[0] Also, here is the manufacturer brochure[1]

[0] http://dangerousprototypes.com/blog/2013/08/21/breakout-boar...

[1] http://www.agilsense.com/useruploads/files/01_HB%20selection...

The difference being Multi-Antenna Synthetic Aperture RADAR vs single antenna RADAR.

Precisely. You need synthetic aperture (or similarly complicated Fourier analysis) with multiple transceive pathways to get a good 3D model of the environment.

It's very impressive that this project managed to miniaturize both the radio hardware and the computational hardware required for this. You need to perform FFTs very fast to make this work.

Many alarm systems have gunn-diode microwave motion sensors in them. Replace this with high-resolution radar and networking for house-wide whole-body gesture recognition.

This might be great for vr? How does it compare to leap motion?

Cool. So when can we get our hands on (or should I say hands off) developer kits?

Another great tool for touch-less interfaces. I think the odd aspect of this Google I/O demonstration is the novelty usage of the technology for a smart watch. This tech seems like it is more difficult than it's worth in that form factor, but for devices that must remain sanitary, say in a kitchen, or a medical device in an OR, I like the potential.

so they are using something like 60-66GHz which beside small antenna size advantage makes for the fastest attenuation http://www.propagation.gatech.edu/ECE6390/project/Fall2012/T... . Pretty logical.

I want to combine one of these chips with a magic mirror.

If this works, then it is one of Google's greatest projects, the number of applications are huge.

Reply on news.ycombinator.com