One of the finalists for the Hats off challenge, Reed Morse, originally wrote that he'd planned to make "a furry hood with eyes. The eyes will have various interactive animations, such as: sparkles based on head movement; changing color based on cardinal direction; pupils that look around; blinking eyelids." (from Adafruit Gemma Hats Off Challenge Competitors ).
I wrote in my first post The hats we decided not to make that I had rejected similar ideas, one (two?) of which I labeled the Wandering Eyes or Emoticon hats. The reason I gave for discarding those ideas was that I didn’t see a way of using NeoPixels to create realistic looking eyes or facial animations within the space of a hat.
I hope I didn't deter Reed from contining the contest, as I've since changed my mind.
At the time, I was only considering the use of NeoPixels, and missed the fact that Gemma can control an 8x8 LED matrix with i2c interface Adafruit Mini 8x8 LED Matrix w/I2C Backpack - Red ID: 870 - $9.95 : Adafruit Industries, Unique & fun DIY electronics an…. It takes 2 wires to control, instead of 1, but the spec says up to 8 different matrices can be controlled on one network.
That makes the modeling of eyes or even a face entirely possible.
Obviously it's too late for me to start building that hat for this contest, and I didn't ever intend to do so, since I didn't want to appear to compete with Reed for the idea he for which he was chosen, However, I couldn’t help thinking about what I would like to try, if I were building that hat. Some of my notes are included here:
First, here are some of my thoughts on the Gemma HatsOff Challenge itself.
I think what drew me to this challenge, and continues to intrigue me about it is the challenge in trying to do as much as possible with as little as possible. The Gemma is a truly limited platform. The processor itself is reasonably powerful, and it has enough memory and program space to provide much more than it’s generally used for. The limiting feature is the I/O – it’s very hard to control much of a system with only 3 I/O lines, even if 1 of them can be used for A/D input and another can be used for PWM output.
I see this contest as similar to the early days of computer usage, when everyone competed to see how much functionality could be coerced from a very limited system.
Today’s systems are generally so powerful that it’s taken for granted that you can use a GPS and 3 axis accelerometer to count the number of steps someone takes during the day. It’s refreshing to be forced to use a less powerful system, and still attempt to get (not useful, but interesting) behavior from it.
OK – enough philosophy. Back to the hat:
Perhaps an accelerometer could be added to the I2C used for control of the LEDs?
That would give feedback on head movement, which could be correlated to eye movement. (Turning the head to the right could result in the eyes looking to the right).
There’s 1 A/D sensor available. What can it do?
Temperature measurement could yield a mood hat, but since I don’t think there’s any actual correlation between a person’s temperature and mood, that’s doesn’t lead to any interesting behavior.
Light sensor– not sure what that would indicate - clouds, movement?
Heartbeat – differences in heart rate can indicate some mood changes (use ear clip?)
Sound – two microphones could indicate the direction of a sound, and the eyes could look in that direction. I'm not sure how to accomplish this with only one input though. Perhaps the difference between two microphones? This could look really cool, as the pinna (external portion of the ear, used to reflect sound back into the sensor) that would help to make this work would also add to the looks of the hat by being what they are (ears!).
I found this paper: http://ai.stanford.edu/~ang/papers/icra09-MonauralLocalization.pdf which appears to show success for a method of sound localization for a single microphone.
Is it possible to use with the ATtiny85’s slow processor and small code space? Probably not. That’s fairly sophisticated computing. The good news is that for the hat to be successful, we don’t need an exact angle to the sound source – simply left or right would be impressive.
Still, lets see what else is available…
This paper: http://www.cnel.ufl.edu/analog/_private/publications/neuromorphic_mic_00638190.pdf suggests the main cue for sound localization is the intensity difference between the two ears. Would a differential amplifier connected to two microphones be able to give a very general localization cue?
Aha! Here’s a schematic for a simple circuit which is reported to work for close sounds. Best of all, it uses pretty simple electronics -- mostly just 4 differential amplifiers, and some passive circuitry (resistors and capacitors) to select amplification values.
http://filear.com/Files/SoundLocalization.pdf
Unfortunately, this device provides two digital inputs (left and right). The Gemma with I2C LEDs has only 1 remaining. Is there a way to combine these two digital inputs into a single analog input?
Yes -- I explored this with the idea for a bicycle map signalling helmet -- two (or more) digital inputs can be read with a single Analog input by connecting them to different resistor networks so that they provide different voltage levels depending on which input is high.
Here's a quick paper calculation to show what I'm talking about:
If this can be made to work even half as well as the original (see the video here http://filear.com/?p=11 ) then a truly impressive hat could be made! Imagine walking around in a crowd with a hat where the eyes turn to look at every source of nearby noise.
I might need to try this one for Halloween next year. My wife and daughter say it would be creepy, but isn't that what Halloween is about?
