This paper introduces a system for gesture based interaction with smart environments. The framework we present connects gesture recognition results with control commands for appliances in a smart home that are accessed through a middleware based on the ISO 24752 standard URC (Universal Remote Console). Gesture recognition is realized by applying three dimensional acceleration sensor information of the WiiMote from Nintendo. This information is trained to a toolkit for gesture recognition that implements machine learning algorithms well known from speech recognition. Our study focuses on two interaction concepts with the aim to exploit the context and special home scenarios. This serves to reduce the number of gestures while in parallel retaining the control complexity on a high level. A user test, also with older persons, compares both concepts and evaluates their efficiency by observing the response times and the subjective impressions of the test persons.