RESUMEN
We present a method to identify the dynamics of a human arm controlled by an implanted functional electrical stimulation neuroprosthesis. The method uses Gaussian process regression to predict shoulder and elbow torques given the shoulder and elbow joint positions and velocities and the electrical stimulation inputs to muscles. We compare the accuracy of torque predictions of nonparametric, semiparametric, and parametric model types. The most accurate of the three model types is a semiparametric Gaussian process model that combines the flexibility of a black box function approximator with the generalization power of a parameterized model. The semiparametric model predicted torques during stimulation of multiple muscles with errors less than 20% of the total muscle torque and passive torque needed to drive the arm. The identified model allows us to define an arbitrary reaching trajectory and approximately determine the muscle stimulations required to drive the arm along that trajectory.
Asunto(s)
Brazo/fisiología , Terapia por Estimulación Eléctrica/métodos , Modelos Biológicos , Modelos Estadísticos , Contracción Muscular/fisiología , Músculo Esquelético/fisiología , Miembros Artificiales , Simulación por Computador , Terapia por Estimulación Eléctrica/instrumentación , Retroalimentación Fisiológica , Humanos , Músculo Esquelético/inervación , Rehabilitación Neurológica/instrumentación , Rehabilitación Neurológica/métodos , Reproducibilidad de los Resultados , Sensibilidad y Especificidad , Terapia Asistida por Computador/instrumentación , Terapia Asistida por Computador/métodosRESUMEN
A major challenge in controlling multiple-input multiple output functional electrical stimulation systems is the large amount of time required to identify a workable system model due to the high dimensionality of the space of inputs. To address this challenge we are exploring optimal methods to sample the input space. In this paper we present two methods for optimally sampling isometric muscle force recruitment curves. One method maximizes the information about the recruitment curve parameters, and the second method minimizes the average variance of the predicted output force. We compared these methods to two previously-used methods in simulation. The simulation model was identified from recruitment data collected during experiments with a human subject with a high spinal cord injury. The optimal sampling methods on average produced estimates of the output force with less error than the two previously-used methods. The optimal sampling methods require fewer system identification experiments to identify models with similar output prediction accuracy.