With the latest advances in technology and the rise of physiological sensors for everyday life, biofeedback is celebrating its revival and is a topic of great interest. The aim of this thesis is a mash-up of biofeedback techniques, modern physiological sensors and 3D technology. It investigates how to create a flexible and reusable biofeedback framework that can be used as extendable platform for future physiological sensors and research projects. It results in a fully operational biofeedback system that can be used to improve body awareness and control. The thesis explains what biofeedback is, investigates physiological sensor modalities and recording techniques, and provides a comprehensive analysis of related work in this domain. Simultaneous acquisition of data from multiple physiological sensors introduces new data management challenges on how to access stored data in an efficient way while still having enough processing power available for data visualization. Rather than just mapping a single value from a sensor like in traditional biofeedback systems, the thesis explains how to create an interactive classification graph, where customizable classifiers combine results from signal processing and map them to one or multiple feedback scores. The thesis extends the traditional biofeedback loop by a control and adjust mechanism and encapsulates analysis and classification from visualization. The two tier architecture allows the creation of state-of-the-art visualizations with any rendering engine. Several sample visualizations are created, including a virtual reality scene using the Oculus Rift in order to investigate the impact of virtual reality in biofeedback. An evaluation with 8 participants, each doing 7 tests, showed that key for successful biofeedback are (1) interaction with a human feedback controller who monitors the session, (2) interaction with a fast responding and simple visualization, and (3) customization of classification. The thesis provides guidelines on how to design useful biofeedback visualizations along with an investigation of the operational capability of physiological sensors and the effect of virtual reality. As a result of this research, a biofeedback framework with a visual and interactive graph-based classification system was created that enables feedback controllers to easily change the classification process and customize it for their users.