Comparação dos Filtros Passa-Alta Butterworth, Hof e ICA para a remoção da Contaminação do ECG no Sinal EMG

Abstract

The electromyography of the trunk muscles seems to be very adequate to study the activity of the abdominal and dorsal muscles during posture control. The characteristics contained in the electromyogram (EMG) signal of such muscles are sources of important information, such as fatigue developed in the muscle during its activation. Unfortunately, the received signals are constantly contaminated by external agents, as in the case of electrocardiogram (ECG) signals, which are capable of hindering the analysis of EMG signal characteristics, resulting in erroneous interpretations. To quantitatively access the differences in both ECG removal and EMG signal preservation, an artificially contaminated EMG signal from peripheral muscles with different patterns and activation levels is used. The ECG used for artificial contamination was recorded by electrodes in a location frequently used by EMG records of the trunk muscles in order to limit the current differences in ECG waveforms in the EMG of trunk muscles. It has been hypothesized that methods requiring separate ECG records are generally superior to methods without the use of such reference. Finally, it is assumed that the independent component analysis (ICA) filtering is more successful than the two alternatives in removing ECG. Three ECG removal techniques were used which do not require the recording of a reference ECG signal. These techniques were: (a) high pass filtering of 30 Hz, (b) sample adaptive filtering, (c) ICA. In addition, two methods for ECG removal including simultaneous recording of the ECG signal were evaluated: the Hof procedure and the ICA filtering. As initial results, four gross signals were obtained, which are two for the biceps muscle activity (right and left), one for the EMG signal collected from the rectus abdominis and one for the ECG signal collected in the sternum. The EMG signals collected in the biceps are divided into 3 activities with duration of 30s each, according to Table 1, adding a total of one and a half minute. The sign of the rectus abdominis and the signal collected in the sternum have only 30 seconds of duration. ICA-based filtering procedures proved to be much superior to the other evaluated procedures. This superiority was due to the fact that ICA and ICA procedures with reference to severely attenuating the ECG noise inserted in the EMG signal, allowed the signal to be reconstructed effectively without the original signal being attenuated too much.