Current mobile devices like mobile phones or personal digital assistantshave become more and more powerful; they already offer features thatonly few users are able to exploit to their whole extent. With anumber of upcoming mobile multimedia applications, ease of use becomesone of the most important aspects. One way to improve usability isto make devices aware of the user’s context, allowing them to adaptto the user instead of forcing the user to adapt to the device. Ourwork is taking this approach one step further by not only reactingto the current context, but also predicting future context, hencemaking the devices proactive. Mobile devices are generally suitedwell for this task because they are typically close to the user evenwhen not actively in use. This allows such devices to monitor theuser context and act accordingly, like automatically muting ringor signal tones when the user is in a meeting or selecting audio,video or text communication depending on the user’s current occupation.This paper presents an architecture that allows mobile devices tocontinuously recognize current and anticipate future user context.The major challenges are that context recognition and predictionshould be embedded in mobile devices with limited resources, thatlearning and adaption should happen on-line without explicit trainingphases and that user intervention should be kept to a minimum withnon-obtrusive user interaction. To accomplish this, the presentedarchitecture consists of four major parts: feature extraction, classification,labeling and prediction. The available sensors provide a multi-dimensional,highly heterogeneous input vector as input to the classificationstep, realized by data clustering. Labeling associates recognizedcontext classes with meaningful names specified by the user, andprediction allows to forecast future user context for proactive behavior.