Evaluation of interpolation methods for EMG arrays

High-density electromyography (HD-EMG) arrays are useful for studying muscle activation in a spatial dimension. This paper studies interpolation techniques for the reconstruction of missing or poor quality channels within an HD-EMG array. Linear, cubic, spline, and nearest-neighbour interpolation methods were evaluated for both 1D (linear) and 2D (array) electrode configurations surrounding a target electrode. The quality of interpolations was measured using the percent residual difference and correlation. For basic 2D electrode configurations, the 2D spline method provided the best results. For 1D configurations, spline interpolation with electrodes selected perpendicular to the muscle fiber was preferable. For sparse electrode arrays, obtained by using every other electrode row/column, the best interpolation method was cubic interpolation over 8 electrodes. Spline interpolation was more sensitive to differences between the EMG from the electrodes used in the interpolation and the EMG in the target electrode; differences became more pronounced in the sparse array configurations. Low interpolation quality was associated with both poor quality channels and regions of transition between high and low activity in the electrode array. Results indicate there is a potential to use interpolation to both identify and reconstruct poor channels in HD-EMG arrays.