Figure 9.

(A) An EMG signal sampled at 2,000 samples/s in the time domain and (B) the power spectrum of the signal in the frequency domain. The signal spectrum contains several spurious peaks. (C) Welch's method breaks the total signal (5 s long, shown in D) into shorter segments (0.5 s) and multiplies (convolves) each segment by a window function (some examples are the Hann, Hamming, and Nuttall windows) before averaging all the modified segments. See Example (viii) in Tutorial Code. (D) In Welch's averaging method, the EMG signal is divided into a number of segments (K). K depends on the length of the segment (L) and the degree of overlap between successive segments, Equation 3 in Supplementary Material. Each successive segment starts D samples after the previous segments. (E) By obtaining an average power spectral density across K segments, the spurious peaks in (B) are reduced. (F) The smoothness of the power spectral density function can be increased by increasing K (i.e., increasing the number of averages, N_AVG), which can be achieved by decreasing the length of L or increasing the overlap between segments. See Example (vii) in Tutorial Code.