Human impedance parameter estimation using artificial neural network for modelling physiotherapist motion

Abstract Physiotherapy (physical therapy) is a form of therapy aimed at regaining patients their bodily limb motor functions. The use of what are called therapeutic exercise robots for such purposes is gradually increasing. Therapeutic exercise robots have been developed for lower and upper limbs. These robots lighten the workload of physiotherapists (PTs) by providing the movements on patients’ relevant limbs. In order to get robots to perform the movements that the PT expects the patient to perform, it is required to determine the mechanical impedance parameters (inertia, stiffness and damping) due to the contact between the PT and patient's limb's, and to ensure that the robot moves according to these parameters. The aim of this study is to estimate these impedance parameters by using artificial neural networks (ANNs). Data from experiments on real subjects were used to train the network, and success was obtained using new data not presented to the network before. Subsequently, the previously acquired output was re-directed to the network with the purpose of developing a network, which can learn more accurately. Results have provided the designed ANN structure can generate necessary impedance parameter value to imitate PT motions.