Analysis of kinematic data in pathological tremor with the Hilbert-Huang transform

This paper presents analysis of kinematic data of tremor patients while performing different tasks with Ensemble Empirical Mode Decomposition (EEMD), a novel noise-assisted data analysis method. EEMD automatically separates raw kinematic data into three components: 1) noise from various sources, 2) tremulous movement, and 3) voluntary movement. Comparison of this technique with other decomposition methods such as recursive forth and back filters or Empirical Mode Decomposition (EMD) shows a better performance; EEMD separation of tremor diminishes EMD error in a 45.2 % (mean error 0.041 ± 0.036 rad/s). Moreover, postprocessing of EEMD separated tremor allows the calculation of the Hilbert spectrum, a high resolution time-energy-frequency distribution that improves analysis of tremors.