First of all, I'd like to thank Sean O'Sullivan for his continued support of accessibility solutions for the Android platform, along with many other exciting projects, through the Rensselaer Center for Open Software.

We're starting out this semester with with a some code for numeric integration interfaced with the Open Intents Android sensor simulator. We'll begin by adding error tollerence to our numeric integration code. Velocity damping and step counting seem to be the best options right now, but our intermediate goal will provide more realistic feedback about the project's viability.

As an intermediate goal, we'll be developing a virtual maze game for the Android. This game will use the same princeples as the final accessibility program, but will not include building maps or directions. It will allow us to optimize our numeric integration methods and investigate using the average step frequency to moderate the numeric integration. Perhaps the game itself will become a useful product in itself.

As a follow up, the Android platform turns out to be uniquely suited to our application. Other platforms (J2ME, the iPhone) have accelerometer support, but other issues would prevent us from working on those platforms. The iPhone has no compass, and although J2ME does support an accelerometer and a compass, we have found that its permission scheme is too restrictive. The Android platform supports an accelerometer and a compass while allowing full privileges to self-signed code.

The compass gives us two options for location tracking on the Android. Step counting is one option, and perhaps we could even reuse our FFT code when keeping track of step frequency. The second option would be a somewhat more straightforward sampling of the acceleration of the phone. We would keep track of the acceleration perpendicular to gravity, get a velocity from that value, and finally we would get a displacement from the velocity. Because accelerometers were not designed for our purposes, we might need to employ both step-counting and numeric integration to keep a somewhat accurate position. Further, it might be necessary to set up Braille 'check points' with simple codes inside buildings, where the user could enter their location manually.

Learning about the G1

This week we have been looking into the new Android phone the T-Mobile G1. It looks like a great device and will be very interesting to work with. It has an accelerometer which also has a compass and can measure yaw and pitch.

We will be using http://code.google.com/p/openintents/wiki/SensorSimulator as a base to work with. I am trying to think of a good way to measure steps. My idea right now is that walking might be some type of periodic motion that we could watch using the acceleration on the three different planes. We will be looking into it more this coming week.