Determining synergy between joint angles during locomotion by radial basis function neural networks

This paper shows that the radial basis function neural networks are suitable tools for determining the synergies between the leg joint angles for cyclic activities. The study was motivated by earlier studies showing the following: 1) cyclic functional movements (e.g., walking) are synergistic [N.A. Bernstein, The co-ordination and regulation of movements, Pergamon Press, Oxford, 1967]; and 2) machine learning techniques for recognizing gait events perform similar when a training set includes one or more joint angles [S. Jonic et al., Three machine learning techniques for automatic determination of rules to control locomotion, IEEE Trans. BME, subm., 1997]. The results of this study prove that the only one joint angle sensor is sufficient to describe a cyclic motor pattern, and that the second joint angle sensor is redundant for cyclic activities, but very useful to detect the change of the mode of locomotion or hazard [D. Popovic et al., Control aspects of active above-knee prosthesis, Intern. J. Man-Machine Studies, vol. 35, p. 751-67, 1991]. The results of the study will be implemented for restoring walking of humans with disabilities using a functional electrical stimulation system.