SYSTEM-IDENTIFICATION OF HUMAN-PERFORMANCE MODELS

The results of an investigation into the application of parametric system identification procedures to human performance are presented. Both stationary and adaptive techniques as well as linear and nonlinear models are discussed. A case study is presented for wheelchair racing that is used to develop multi-input/single output models. The need for model order reduction and methods for quantizing training data are also discussed. The results suggest that nonlinear autoregressive moving average with exogenous (Narmax) models are better predictors of performance than the other models investigated. In addition, the model developed for wheelchair racing suggests that motivation may play a more important role in predicting uncharacteristic performance in elite athletes than aerobic or strength training.