Algorithm 1: Overall fatigue level computation
Input: Raw sEMG and corresponding acceleration signals
Output: Overall fatigue level
1:	Filter and resample the input signals;
2:	For i from 1 to Nm
3:	Calculate the self-covariance of the filtered acceleration signal ϕACC;
4:	If ϕACC < ε, the movement is recognized as a periodic movement. Then, the corresponding sEMG signal is segmented and connected to form a new sEMG signal, otherwise, do nothing;
5:	Compute the initial mean frequency of the sEMG signal fmean,0;
6:	Initialize λ, φ̂t and pt. φ̂t(0) can be 0 or an initial guess of the slope parameter, as pt is defined as t−2, pt(0) has to be set as a relatively large number;
7:	Compute fmean,i and update φ̂t and pt by Equations (11) and (12);
8:	Calculate the localized fatigue level lfatigue,i by Equation (9);
9:	End For
10:	Compute the overall fatigue level lfatigue by lfatigue,i (Equation (13)).