An optimized method for tremor detection and temporal tracking through repeated second order moment calculations on the surface EMG signal.

In this study, the problem of detecting and tracking tremor from the surface myoelectric signal is addressed. A method based on the calculation of a Second Order Moment Function (SOMF) inside a window W sliding over the sEMG signal is here presented. An analytical formulation of the detector allows the extraction of the optimal parameters characterizing the algorithm. Performance of the optimized method is assessed on a set of synthetic tremor sEMG signals in terms of sensitivity, precision and accuracy through the use of a properly defined cost function able to explain the overall detector performance. The obtained results are compared to those emerging from the application of optimized versions of traditional detection techniques. Once tested on a database of synthetic tremor sEMG data, a quantitative assessment of the SOMF algorithm performance is carried out on experimental tremor sEMG signals recorded from two patients affected by Essential Tremor and from two patients affected by Parkinson's Disease. The SOMF algorithm outperforms the traditional techniques both in detecting (sensitivity and positive predictive value >99% for SNR higher than 3dB) and in estimating timings of muscular tremor bursts (bias and standard deviation on the estimation of the onset and offset time instants lower than 8ms). Its independence from the SNR level and its low computational cost make it suitable for real-time implementation and clinical use.