Tikalon Blog by Dev Gualtieri

Tikalon Header

Gesture Recognition

June 17, 2013

At the start of my undergraduate education , I briefly attended a college whose emphasis was on engineering . This was a mistake for a physics major, so I left a short time later. I did, however, gain some valuable extracurricular knowledge about radio broadcasting . Since it was a select school, I met quite a few creative and interesting people. The school was so select, in fact, that fully half the students were above average .[1]

The conversations at group lunches were highly technical. We would often invent useful devices, at least on paper - napkin paper, that is. One of these was a gesture control system in which waving an unaided hand would control a device such as a television . Nowadays, there are quite a few technologies to tackle this problem, the most effective being machine vision systems , as used in the Microsoft Kinect device.

In the late 1960s, however, our technology choices were limited, but there were two that would have enabled such a device. The first is radar , since Doppler signals from a waving hand would be easy to detect. The radar generator would have been something like a klystron , although solid state microwave generators, such as the IMPATT diodes , tunnel diodes were just becoming available.

Cutaway of a Varian V-260 reflex klystron

Cutaway of a Varian V-260 reflex klystron . Such devices attached to rectangular waveguides , usually at the X band .

(Cutaway model provided by Walter Z. Seldon and photographed by Erbade, via Wikimedia Commons.)

Another, more easily built device would be based on the Theremin . I mentioned Léon Theremin , the inventor of the Theremin, in a previous article (Cyber Warfare, November 10, 2011). The Theremin uses the electrically conducting human body as a plate in the variable capacitance of a radio frequency oscillator ; so, proximity of a body part, such as a hand, will cause a detectable frequency change.

We sit in our offices and living rooms immersed in electromagnetic fields , some of which are intentional, and some are not. The most useful is the light that comes from our lamps and display screens . The most annoying is the power line emission that gives us the buzz on our AM radios and stereo music systems .

Environmental fields can serve as the basis for object detection. One interesting example of this was the use of cosmic rays by Nobel Physics Laureate , Luis Alvarez , in a search for hidden chambers in an Egyptian pyramid in the 1960s. Scientists in Mexico have applied this same technique to the pyramid at Teotihuacan .[2] Recently, more primitive methods have been used at the Nohmul complex in Belize .[3]

The Pyramids at Gizah

The Pyramids at Gizah in Egypt . During my college years in the 1960s, many New Agers believed that the pyramid shape focused "cosmic energy."

(Photograph by Ricardo Liberato , via Wikimedia Commons.)

One type of electromagnetic radiation pervading our home and office environments is Wi-Fi , which is a physical layer in local area networking . Tablet computers , laptop computers , desktop computers , game consoles , computer peripherals and smartphones all have Wi-Fi capability. It sure beats snaking Ethernet cable throughout a house, as I did many years ago.

University of Washington computer scientists have decided to harness this sea of Wi-Fi radio waves as a means for gesture recognition. The Washington research team includes Shwetak Patel and Shyam Gollakota , both assistant professors of Computer Science and Engineering , and doctoral student , Sidhant Gupta . Their technology, which they call "WiSee," will be presented at The 19th Annual International Conference on Mobile Computing and Networking later this year, but details are available online.[4-7]

Figure caption

A display showing the change in wireless signal strength across frequency channels in real time as a user moves his hand. (University of Washington image.)

One advantage of this method is that Wi-Fi signals travel through walls, so gesture detection is not limited by the line-of-sight restriction of camera systems. The method is based on the Doppler effect, which in the gesture case is just a few hertz . The entire Wi-Fi spectrum is scanned, looking for subtle changes in signal strength as a function of frequency .[4]

At this point, the WiSee system can identify nine different whole-body gestures, such as pushing, pulling and punching. In a test environment of five users in a two-bedroom apartment and an office environment, 94% of 900 gestures were accurately detected.[4] As for security , a specific sequence of gestures can be used as a password .[4] My college classmates and I were just a few decades too far ahead of the technology curve on this one.

References: