How To Make (Almost) Anything

Anmol Madan, Human Dynamics Group, anmol@media.mit.edu

Welcome to my web page devoted to MAS 863.04 ! This page started out as a quick hack, but is slowly evolving into a detailed representation of my work in the class. ;-) Almost all pictures were taken using my Nokia 6600 cell phone. 

This page shows the evolution of my final project, InterestNetworks. Click here to see my weekly assignments.

I had an  idea of the direction of my final project, although the details were quite hazy. 

For my final class project, I wanted to  build the next-generation wearable platform for my research group that would support real-time speech feature, physiology and motion processing. The system core is the Sharp Zaurus SL6000 PDA or the Motorola A760 Linux cell phone, both of which support  Bluetooth. I wanted to design a bluetooth accelerometer and integrate it with a bluetooth GSR module, designed by the Affective computing group. Part of the problem was exploring innovative techniques of packaging all of these into a cohesive, intuitive system and user experience. 

Applications

This new-generation platform will take over from the Mithril platform as our primary research infrastructure. I plan to use this system  towards my Master's thesis, by real-time behavior modeling to predict interest in various applications. We are conducting pilot studies in movie audiences, negotiations, speed dating, brainstorming, and other constrained scenarios where this multi-modal, wireless system will be rapidly deployed. In addition, I expect that this system will also be used by other members of my group for applications in healthcare, social networking and group coordination.

More detailed information about these applications and real-time machine learning models used is available in these papers. 

Prior Art

Mithril, MIT Media Lab
Our current wearable system for motion, speech and physiology, that uses wired sensors with the Sharp Zaurus PDA.

Bluetooth Accelerometer, Contextual Computing Group, Georgia Tech
A Bluetooth accelerometer based on a similar Taiyo Yuden module

GNOMES, Rice University
Bluetooth sensors used for sensing environmental conditions

QBIC, ETH Zurich
Bluetooth enabled, StrongArm based wearable computer

Refining the Idea

The bluetooth accelerometer would be based around the Atmel AVR . I planned to use two orthogonally arranged ADXL202  2g MEMS accelerometers and a Taiyo Yuden EYMFCAMM-XV bluetooth module. I used the ATmega32, which could support some basic speech processing (speaking/impulse noise threshold detection) in the future if needed. I'd also be breaking out the remaining AVR pins so that we can easily attach a GSR or simple speech processing daughter board. The system would be packaged in  re-sealable plastic casing (either vacuum-formed or laser-cut)  so we can periodically change coin-cell  or Li-Poly batteries, while maintaining durability.  

I also thought it would be fun to play with the way GSR leads are packaged. I saw an image of a BodyMedia  EEG art exhibition in Japan, which inspired the GSR packaging above. I though it would be cool to cast it in silicone.  The image with the hand shows a clay rendition of a proposed packaging scheme.

Parts and Materials:

AVR ATmega32 micro controller
ADXL202 JE accelerometers, 2
TAIYO YUDEN EYMFCAMM-XV bluetooth module
Assorted RC components, regulators, headers
Spin etching materials
Vacuum forming PET

Bluetooth GSR by Affective Computing
Standard GSR leads and stickies
3D Rapid Prototyping ABS Plastic
Silly Smooth RTV casting Silicone
Laser cut mold

Eagle Design Files: Blueaccel.sch and Blueaccel.brd

Final Presentation

This is what version 1.0 beta of the  bluetooth accelerometer looks like. The large triangular board has the ATmega32 and an ADXL. The smaller board has the other ADXL, and will be mounted perpendicular to the first board. The accelerometer circuits are designed to measure up to 50Hz (human motion is typically less than 10Hz), each channel sampled 100 times a second by a 4Mhz AVR. You can use this information to calculate exact component values here.  The bluetooth module is operating in auto slave mode, and waits for a (pin less) connection to  start sending data. Data is sent from the uC to the bluetooth transceiver using the UART at 115kbps.

In order to cast the silicone, I laser-cut an acrylic mold. For version 1.0, I used blue RTV silicone and press-fit connectors. When I did the silicone casting, I couldn't find a tutorial online, so I decided to write my own silicone casting tutorial. (Thanks Dick!). 

This looked okay and worked great, but I thought translucent silicone would be cooler.  I made version 2.0, which had squeeze-ease type of a connector. In order to correctly create a mold, I put a set-screw in the grip while casting, which I later removed. This basically meant that you has to squeeze the silicone to lock the connector on a leads - fun!

For the GSR module, I also 3D printed a custom-sized packaging on the Stata, in ABS. 

The final system, streaming data  looked like this.