Comparison of spatial filters for HD-SEMG single motor unit feedback

McNaught, Andrew and Cescon, Corrado and Vieira, Taian M. M. and Lester, John and Merletti, Roberto (2010) Comparison of spatial filters for HD-SEMG single motor unit feedback. In: Proceedings of the VII Motor Control Conference - MCC, 24-27 Sept 2010, Varna, Bulgaria.

[img] Slideshow
AM McNaught - MCC 2010 - Spatial Filter Presentation.pdf - Presentation

Download (4095Kb)


AIMS. To compare the ability of various spatial filters to enhance the amplitude of a single motor unit’s action potential from the background surface electromyographic (sEMG) signal with the aim of identifying the best spatial filter to use for real-time HD-sEMG feedback of single motor units. Spatial filtering is well suited to real-time applications due to computation simplicity. Single motor unit (MU) enhancement would allow the use of simple MU detection methods, such as threshold techniques, to allow real-time feedback of MU features. METHODS. Twelve healthy male volunteers (age, mean ± S.D., 26.75 ± 3.96 years, height, 1.8 ± 0.07 m, weight, 76.17 ± 10.32 kg) participated in the study. The right hand was secured in a brace, allowing isometric contractions of the abductor pollicis brevis (APB) muscle (Figure 1a). Subjects performed contractions at levels of 1, 2, 5, 10 and 20% of the maximum voluntary contraction (MVC) level, measured with a load cell. Monopolar sEMG signals were acquired from the APB muscle with a 64-channel dry electrode grid (Figure 1b) and a multichannel EMG amplifier (EMG-USB, OT Bioelettronica). Signals were decomposed off-line using the Convolution Kernel Compensation (CKC) algorithm (Holobar and Zazula 2007) and averaged in order to extract the templates of each motor unit (MU) for each channel. The following spatial filters were then applied to the MU templates: Longitudinal Single Differential (LSD), Longitudinal Double Differential (LDD), Transversal Single Differential (TSD), Transversal Double Differential (TDD), Bipolar Transversal Double Differential (BiTDD), Normal Double Differential (NDD) and second order Inverse Binomial (IB2). Spatial filters were compared in terms of a Selectivity Index (SI); which depends on the amplitude ratio between the largest (PMU1) and second largest MU (PMU2) of each filter (SI = 1 – PMU2 / PMU1), where P is the absolute peak value. This gave an indication of each filter’s ability to highlight the largest MU from the other MUs after filtering. The two largest MUs (PMU1 and PMU2) for each location of the LSD and NDD filters are shown in Figure 1c. RESULTS. The NDD filter showed the highest SI amongst all investigated filters, followed by LDD. The worst performing filter was the monopolar. SI for each filter seems to depended on the location as can be seen in Figure 1d), maps of SI values interpolated across all the channels for the LSD and NDD filters. CONCLUSIONS. From the results it seems that certain spatial filters have the ability to highlight different single MUs over certain regions of the muscle, indicating that there may be a spatial filter that is better suited for enhancement of the amplitude of a single MU from other MUs contributing to the sEMG signal. ACKNOWLEDGEMENTS. This research was supported by funding from the National Research Foundation (NRF) of South Africa and from Compagnia di San Paolo and Fondazione CRT, Italy.

Actions (login required)

View Item View Item