Skip to Main Content
To investigate the validity of common spectral shift parameterizations, two closely related studies were conducted. Simple linear and exponential regression models were applied to the time-series myoelectric mean power frequency and median power frequency data of two normal subjects as well as a single subject over multiple trials. Statistical analyses were applied to test the hypothesis that no significant differences in extracted features (measurands) exist between different epochs of spectral parameter data. Statistically significant differences were found to exist in both studies, implying that the trial duration can significantly influence the obtained measurement values, and that neither linear nor simple exponential regression models adequately characterize the time-dependent changes. It is concluded that parameterization techniques based on statistical curve-fitting strategies should be supplemented with physiologic models to gain insight into the functional nature of the time dependency. Results are discussed in this context.