EMG-Based Characterization of Pathological Tremor Using the Iterated Hilbert Transform

The identification and characterization of pathological tremor are necessary for the development of techniques for tremor suppression, for example, based on functional electrical stimulation. For this purpose, the amplitude and phase characteristics of the tremor signal should be estimated by effective detection techniques, either from the kinematics or from muscle recordings. This paper presents an approach for the estimation of the characteristics of pathological tremor from the surface electromyogram (EMG) signal based on the iterated Hilbert transform (IHT). It is shown that the IHT allows an asymptotically exact modeling of the tremor and the voluntary activity components in the surface EMG, and an effective demodulation of the pathological tremor parameters. The method was tested on signals generated by a recent model for tremor generation as well as experimentally recorded from patients affected by pathological tremor. The results showed the ability of the proposed approach to demodulate effectively the tremor amplitude (average correlation with imposed amplitude: R² = 0.52), the frequency (root mean square error in frequency estimation: 2.6 Hz), and phase, as well as the degree of voluntary activity (correlation with simulated inertial load: R² = 0.62 ). The application of the method to the experimental data indicated that the estimated tremor component closely resembles inertial measurements of limb movement (peak cross correlation across four patients: 0.62 ± 0.15). Compared to the performance of empirical mode decomposition, the proposed method proved to be more accurate for tremor characterization without a priori knowledge of the tremor characteristics. This method can be used as a part of a control system in strategies for suppression of tremor.