Real-time registration of body segment angles is essential in artificial body position control. A new method is presented for the real-time calculation of the lower extremity angles using data obtained from pairs of two one-dimensional accelerometers. It is shown that, assuming rigid-body dynamics and simple hinge joints, relative angles (i.e. angles between segments) can be calculated without integration, thereby solving the problem of integration drift normally associated with accelerometry. During the stance phase of walking, the relative angles can be transformed to absolute angles (i.e. relative to the gravitational field direction) for the different leg segments. The feasibility of relative angle calculation is demonstrated by calculation of the knee angle of a healthy subject. Stability and resolution were demonstrated with measurements during standing. Measurements during standing up, sitting down and walking showed that shock (heel-strike) and skin movements, due to movements of the underlying muscle tissue, are the main error sources. Additional signal processing, e.g. low-pass filtering, can be used to diminish this error. The accuracy of the knee angle found is shown to be high enough to be used in a feedback controller for functional electrostimulation of the lower extremities.