Characterization of vibration-induced artefacts in EMG signals

Hand amputation due to traumatic events or diseases causes relevant impairments on the capabilities of a person. The use of hand prostheses has the potential to radically improve the quality of life of hand amputees. Active hand prostheses are usually controlled by myoelectric signals collected by means of surface electrodes placed on the residual stump muscles. These signals are used to identify movement intention and control hand fingers. Vibrotactile stimuli have been proposed to provide the user with a feedback aimed at improving the usability of the prostheses. Usually, vibrator and bipolar electrodes are placed in close proximity for practical and constructive reasons. However, in this configuration, vibrations could cause a movement artifact on the sEMG signals used to control the hand movement. This study focused on the characterization of the vibration artifact on sEMG signal. Three potential sources of the motion artifact were analyzed: the changes of the impedance unbalance, the common-mode and the differential-mode variations of the half-cell potential. The effects of these factors were studied in three steps: theoretically; using custom-made circuits and experimentally. Initially, a bipolar EMG detection system was built on an electronic board. The circuit included a Gold Standard front-end electronics (INA333) and the electrical models of the electrode-skin contact made with discrete components; RC parallels for the impedances (Ze) and batteries for the half cell potentials (Ehc). Variations of the Ze and Ehc simulating the three hypothesized causes of motion artifact were obtained through switches. With this approach, we obtained a model-based characterization of the movement artifact that was compared with the theoretical analysis of the considered circuits. In the second step, the tests of the first phase were repeated by replacing the circuital model of the electrode-skin contact with actual electrodes on the subject. Movement-induced changes in the properties of electrode-skin contact were simulated by changing the size of the exploring and reference electrodes. In the third step, we applied a motor-driven vibrator at different distances from the exploring and the reference electrodes to induce motion artifacts in bipolar EMG recording in the conditions previously simulated. The theoretical considerations and the analysis of the recorded signals showed that the common-mode variations of the half-cell potential cause the appearance of spike-like artifacts due to both the electrode-skin impedance unbalance and the input capacitance of the front-end amplifier. As expected, the differential mode changes of the half-cell potential induce shifts of the baseline of the signal, and the impedance unbalance causes the modulation of the powerline interference. These observations were confirmed by the experimental analysis (step three). These results demonstrate the importance of the proper skin preparation, of the positioning of the reference electrode (i.e. far away from the vibration location) and of the design of the front-end amplifier. Moreover, our findings suggest the potential usefulness of the two-electrodes biopotential amplifiers in case of applications where high robustness to vibration artefacts is required.