SS05 - Bachelor Praktikum for Human Computer Systems

In Wearable Computing it is the goal that the computer perceives the world from a user's perspective -- "seeing what he sees, hearing what he hears and feeling what he feels". Based on sensor data it can infer the user's context, such as the user's current activity ("walking", "riding bike", "shopping"), estimating his current location or even predicting what he will do next.

One way to obtain annotated data for classifier training is to equip users with sensors, record data and annotate it afterwards. These annotations, together with the sensor data, can later be used as training data for machine learning models.

There is an existing tool developed at the MIT Media Lab, but this tool has some flaws that cannot be remedied because it is closed source. Therefore the task is to reimplement that tool. Some characteristics of the application should be:

• graphical user interface following HCI principles
• handling of video, audio, acceleration and other types of sensor data
• extensibility of the tool (new data sources)
• reading/writing a file format for annotations that is based on an open standard such as XML (the format will be provided by us)
• efficient handling of large volumes of sensor data (caching, generation of index files)
• synchronized display of different input media
• an interface to Matlab for postprocessing of journal data