Enhancing the performance of upper limb gesture reconstruction through sensory fusion

A novel method devoted to the reconstruction of the joint angles in a kinematic chain is described. The reconstruction algorithm is based on the fusion of the information deriving from inertial sensors (accelerometers) and conductive elastomer strain sensors. Accelerometers provide a reliable reconstruction when they are employed as inclinometers in quasi-static conditions. They suffer from artifacts when they are used to detect fast movements or when interactions with the environment occur. The knowledge of the frequency components of the movement to be detected permits removal of these artifacts. Conversely, conductive elastomer sensors have a complex dynamic response, but they can easily provide the frequency content of the movement to be detected. A filtering strategy of the inertial sensor signals based on the elastomer sensor response provides a reliable reconstruction of joint variables during the movement.