Characterisation of Physiological Tremor using Multivariate Empirical Mode Decomposition and Hilbert Transform

Fatigue-induced physiological tremor (FIPT) is undesirable when performing micromanipulation tasks that require high precision. It is important to characterise this form of tremor to aid in identifying and suppressing it from the intended micromanipulation task. Researchers have used surface electromyography (sEMG) and mechanomyography (MMG) separately or in combination to study tremor, which is further processed using Fourier transform-based techniques. The major drawback of using these techniques is that it assumes the signals are linear and stationary. On the contrary, Empirical Mode Decomposition (EMD) can provide localised information on energy of the non-linear and non-stationary signals at a particular time and frequency and hence is considered superior to the Fourier transform-based techniques when analysing signals like physiological tremor. This paper characterises physiological tremor by extracting the frequency band of interest using multivariate empirical mode decomposition (MEMD). The extracted frequency band is assessed using Hilbert spectral analysis for energy estimation. Energy Ratio (ER) is the parameter proposed in this study to indicate fatigue-induced physiological tremor. The linear regression of the ratio across task epoch (TE) showed an increasing trend with R²≈0.7 for sEMG signals and R²≈0.9 for accelerometer signals to indicate levels of fatigue increase.Clinical Relevance - This study presents an effective & versatile indicator of FIPT. The characterisation method discussed in this paper will form an integral part of creating control strategies that can eliminate undesired consequences of fatigue-induced tremor in prolonged surgical manipulation. In addition, in surgical training modules, it aids the learning rate of novice surgeons.