Reducing cross terms effects in the Choi-Williams transform of mioelectric signals.

This study aims at investigating the effect of removing the negative values of Choi-Wiliams distribution (CWD) related to the electromyogram (EMG) for visualization and instantaneous median frequency (IMF) estimation. Beyond the EMG signals from triceps surae and biceps brachialis, the CWD was applied in a simulated sinusoidal signal as like in stationary and non-stationary simulated EMG signals (SES). The CWD negative values of all simulated and EMG signals were removed. The IMF values were obtained for SES and EMG. The CWD IMF values from SES and EMG were thus compared with the IMF values from short time Fourier transform (STFT) by means of correlation. The suppression of negative values from the CWD reduced cross terms influence and improved visualization, as shown by the increased correlation coefficient between the IMF values. Before this suppression, the extracted IMF values showed large oscillation along the time, with various spurious values beyond 500Hz, which disappeared after the suppression. Moreover, this procedure seems to be especially useful for non-stationary signals.