Saskia Baltrusch

94 Chapter 4 metabolic rate from the total metabolic rate during walking and lifting. Net metabolic cost for walking was obtained by normalizing net metabolic energy expenditure to walking speed and was expressed in J/kg/m. For lifting, net metabolic cost was not normalized to speed and was expressed in J/kg/s. EMG signals were high-pass filtered with a 2 nd order Butterworth filter (cut-off frequency 20Hz, bidirectional) to remove movement artifacts. Additionally, a 4 th order Butterworth band stop filter (49Hz – 51Hz, bidirectional) was applied to remove hum. After rectifying the data, the data were filtered with a 4 th order low pass Butterworth filter (cut-off frequency 4Hz) to create a linear envelope. Next, we normalized muscle activity for each muscle to the maximum of the linear envelope obtained in the MVC trials. EMG envelopes were then normalized to cycle time and averaged over cycles. Kinematic data were filtered with a 2 nd order low pass Butterworth filter with a cut-off frequency of 5Hz. For the lifting trials, the instantaneous 3D knee, hip, trunk and lumbar joint angles were calculated [24]. We only analyzed angles in the sagittal plane of the anatomic reference frames, hence: knee flexion/ extension, hip flexion/extension, trunk inclination and lumbar flexion/extension. Knee peak angles were used to define a lifting cycle. One lifting cycle was defined as bending down, lifting the box, holding the box in upright position, lowering the box, and returning to upright position without the box again. Range of motion per joint was calculated and averaged over movement cycles, generating an average value for each condition per participant. The centers of mass of all segments were used to calculate the total body center of mass. By taking the mean of the vertical distance travelled by the body center of mass over each separate lifting cycle, we arrived at the average range of motion of the body center of mass during a lifting cycle. For the walking trials, heel strikes for each cycle were determined, using the data of the two heel markers. Left and right heel strikes were defined as the instants of local maxima in the horizontal position of the respective heel marker in the sagittal plane. Heel strikes were plotted and visually checked for detection errors. The heel strikes were used to calculate average stride times and stride lengths per trial. Stride times were calculated as the time differences between subsequent left heel strikes. Stride lengths were calculated as the distances