Turning anything into a touch sensor

This year at the CHI conference in Austin, [Munehiko Sato], [Ivan Poupyrev], and [Chris Harrison] out of the Disney research lab in Pittsburgh demonstrated their way to make touch sensors out of anything. Not only to they suggest using the surface of your skin to control cell phones and MP3 players, they’re also able to recognize touch gestures, like poking or grasping an object. That sounds a little heady, so check out the video of the Touché tech in action.

Like the capacitive touch sensors in our phones and tablets, Touché measures the rise and fall of a capacitor’s charge over time. Unlike  other touch sensors, Touché scans the capacitor at different rates, allowing for a ‘capacitive profile’ that is used to recognized touch gestures.

The applications for this tech are nearly innumerable; the team demonstrated scolding someone for eating cereal with chopsticks (yeah, we know…), an on-body music player interface, and gestures for an office doorknob that notifies passersby if you’ve stepped out for a minute or are gone for the day.

It’s a very interesting build, and we give it two weeks until someone replicates this build. We’ll be sure to post it then.

24 thoughts on “Turning anything into a touch sensor

  1. I wonder what the minimum hardware requirements for something like this would be; I have poor understanding of capacitive sensing, but it seems you would need decently quick hardware to be able to sample several frequencies at a decent resolution. There are quiet a few applications I can think of for using this in a vehicle…

  2. This is great, looking forward to someone writing a software library for this for phones. Would you have to train this for each person though as we are all have a little different capacitance profile?

  3. Cool, though the thing about capacitive touch sensing (much like a lot of semiconductor work, or so I hear) seems to always be that everything you thought was was new or innovative was already thought up of in the 80s (or patented, and already expired). The technology to do things like this was easily out there–it’s wasn’t impossible to do a lot of the capacitive touch sensing shown today with a bunch of logic gate. It was just prohibitively expensive for most applications and prohibitively expensive to experiment and design good capacitive sensing systems. Today it’s pretty trivial to play around with and optimize capacitive touch sensing.

    Anyway, I see a bunch of pretty cool graphs which can tell you a lot about what’s going on, but for a final application I really fail to see why the entire spectrum is necessary. For most applications, I’d imagine you could pull off quite a bit with one or two frequencies. More of a thing is that having done some work on these sensors in the past, and having read a lot about practical implementation issues with capacitive sensors, I highly doubt that most of those sensing profiles are robust to humidity, environmental noise, and different geometries. Probably took a couple of takes and a few trained grad students to get something that looked good. . .

    1. Considering the base level of hardware needed, would it not be trivial to also gather humidity and temperature data? Could a neural network, such as the one used in the handwriting recognition post from a few days ago, increase accuracy? To address the environmental noise, could you use the signal difference between the primary input and a secondary capacitive sensor?

  4. Personally, I’m very skeptical about hand-capacitance control devices.

    At one point, my home was fitted with hand-capacitance wall switches, which allowed you to light and dim ceiling fixtures with the brush of your fingers.

    Pretty sweet… until the first monsoon storms came. Every time a bolt of lightning struck nearby, the lights throughout the house would come on, go out, or change intensity. This scared the heck out of a house-sitter who stayed alone in our home one time when we went on vacation. She thought the place was haunted.

    Speaking of touch control, Delta or a similar manufacturer advertises an electronic faucet on television, which turns on the water if you touch the faucet spout with the back of your hand. I cringe every time I see it. It would be just wonderful to come home from work and find a few thousand gallons of water on the floor because an afternoon rain storm activated a faucet in the kitchen sink.

    Give me toggle switch with a bat handle on it, or a faucet with a valve handle.

    1. Faucets with IR-sensors work really well and require no physical contact, so the are very hygienic. I believe they also somehow generate the needed power from the flow of water.

  5. I see errors based on humidity changes. the table for example has slight difference between changes, what if I wiped the table? overall capacitance will change while it is wet for the next 10-30 minutes depending on humidity.

    the touchscreen uses a wide range and difference but is still freaked out by moisture.

  6. from the description the hardware seems to be extreme over kill
    a microprocessor and a wave generator?
    any stamp or chip will have a couple bits of i and o
    TC=R*C
    1 ohm 1 farad = 1sec
    actually 1 second will take it to 63%of the applied voltage the second will remove 63% of the difference so 10seconds is 99.99 * of the applied voltage

    parallex stamp says
    3.3v out
    1.6 volt in logic 1
    6 picofarad input loading

    so if you connect an output pin to a 1megohm resistor to a wire to the sensing plate
    and run the wire to the inputpin
    it will take 6 microseconds to trigger the input pin with a short wire
    fet input so only added resistance will be the socket

    the lo end stamp processor im looking at has a 0 to 100Mhz vco onboard so it a bit more than needed too

    the cpu has a 12megaHz clock so you need to write a program

    mark the time (or reset)
    set the outpin high check the input in to see when it goes low
    when it goes low set outpin low and see how long it takes for the input pin to go low
    mark the time (time elapased for hi/low)
    this time is the capacitive load of the circuit (and whatever you hook the wire to)

    this is several instructions in a loop
    so with a 12MHz cpu and minimal capacitance (wire length) you should have an answer in about 6 microseconds
    a 10Picofarad cap (about the smallest discrete ive seen ) should take about 20milliseconds to charge/discharge

    if you touch the wire (or get close) it will take longer to change
    if you let go it will be back to the original time constant

    if you add a numeric display you will have a rather good capacitor tester

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s