Saskia Baltrusch

187 Chapter 7 Electromyography EMG data were filtered using a 4th order Butterworth band stop filter between 49 and 51 Hz to remove power line hum. Subsequently, the data were high-pass filtered (2 nd order, 20 Hz), rectified and low-pass filtered (4 th order, 2.5 Hz). Next, we normalized the EMG data to the maximal amplitude of the signal obtained in the MVC trials and to cycle time. The normalized data were averaged over both body sides and over cycles. 2.6 Statistics To test for statistically significant differences between control condition and exoskeleton condition, we conducted paired t-tests for all the outcome variables. Critical level of significance was set to Alpha = 0.05. After a first inspection, two participants showed errors caused by movement of the markers or loosening of the EMG electrodes. We therefore did not include these trials in the analysis of the kinetic, kinematic and EMG data. The number of participants included in the statistical analyses are therefore different for different dependent variables and are reported for each outcome. 3 Results 3.1 Metabolic cost Wearing the exoskeleton decreased net metabolic cost of lifting by 18 % (means (sd): 5.63 W/kg (1.26) vs. 4.64 W/kg (1.38); p=0.000) (Figure 4). One participant showed an increase in metabolic cost when wearing the exoskeleton. This could be explained by the weight of this participant (120kg), and the resulting lower relative effect of the exoskeleton, compared to the remaining participants. 3.2 Kinematics and mechanical joint work We did not find a significant effect of wearing the exoskeleton on peak angles in knee flexion, hip flexion, lumbar flexion and trunk inclination (Figure 5). All participants adopted a squat or semi-squat lifting style in the control and exoskeleton conditions. 7