Fatigue Evaluation through Machine Learning and a Global Fatigue Descriptor
Table 2
List of the various EMG and HRV fatigue descriptors with the typical evolution trend in fatigue conditions and the most appropriate combination of window size and time-step.
Signal
Domain
Parameter
Window size
Time-step
Coefficient of variation
Half length of 95% CI
Typical evolution
EMG
Frequency
Median frequency
10 muscular activations
1 muscular activation
EMG
Time frequency
Median frequency
10 muscular activations
1 muscular activation
EMG
Time frequency
Major frequency
10 muscular activations
1 muscular activation
EMG
Time frequency
Major time
25 muscular activations
0 muscular activations
HRV
Time
Maximum duration of RR intervals
30 s
3 s
HRV
Time
Minimum duration of RR intervals
30 s
3 s
HRV
Time
Average duration of RR intervals
30 s
3 s
HRV
Time
rmsSD
50 s
45 s
HRV
Time
SDNN
30 s
3 s
HRV
Time
Triangular index
60 s
30 s
HRV
Time
SD2
30 s
3 s
HRV
Frequency
Power in LF band
30 s
3 s
HRV
Frequency
Power in HF band
30 s
3 s
HRV
Frequency
Median frequency
30 s
3 s
refers to the combined slope and CI to the confidence interval ( half length of 95% CI). SDNN refers to standard deviation of NN intervals; rmsSD refers to root mean square of successive differences; Poincaré Standard Deviation/Dispersion of points perpendicular (SD1) or along (SD2) the axis of line-of-identity (ellipse semiaxes 1 and 2). Low-frequency band ([0.040; 0.150] Hz); high-frequency band ([0.150; 0.400] Hz).
