The “Tiny Task Network Titan” is a framework for gesture and motion recognition. Titan is developed at the Institute for Wearable Computing at ETH Zurich.

“Gathering and Labeling Sensor Data on a Mobile Phone with TITAN” is my bachelor thesis that I wrote together with Jonas Huber.

In order to make the Tiny Task Network Titan usable for gathering and labeling of sensor data, we enhanced it by several aspects. Titan can now use Bluetooth sensors, synchronize and merge data streams, do GUI-based online labeling and write both, labeling and data, to text files. In addition, an example of an online activity recognition on the mobile phone was implemented. The performance of this extended Titan framework and the activity recognition in its entirety have thoroughly been tested on the HTC Touch HD mobile phone with Windows Mobile 6 as well as on a desktop computer. On this mobile phone, the maximum feasible configuration consists of two sensors running at a sampling frequency of 64Hz, while this limitation does not hold for a desktop computer. The battery life of the mobile phone while running Titan and using Bluetooth sensors accounts for over seven hours in all tested cases. Secondary, an evaluation of the most common mobile operating systems showed that the future development of Titan should focus on Android or Maemo.

The thesis can be found at http://www.wearable.ethz.ch/education/sada/autumn2009/GatheringWithTITAN and here.

Sourcecode

The source code can be found at the following two locations:

Additionally, a patch of all my changes can be found here: Gathering_and_Labeling_Sensor_Data_on_a_Mobile_Phone_with_TITAN.patch