Poster abstract: Gesture recognition via continuous maximum entropy training on accelerometer data

In this work, we present a gesture recognition system via continuous MaxEnt training on accelerometer data. We are working on embedding feature selection into MaxEnt model training to more intelligently narrow the feature set. We are also working to simplify the modeling process so developers can more easily build sensor-based applications. Finally, we are exploring the use of MaxEnt to model higher-level activities.