*2.3. Equipment and Data Analysis*

To collect inertial sensor data, two wearable IMUs; (APDM, Inc., Portland, OR, USA) were attached to the sternum and forehead of the participant. Each IMU included a tri-axial accelerometer (±6 g), gyroscope (±2000 ◦/s) and magnetometer (±6 gauss) that measured at a sampling frequency of 128 Hz. Moveo application (APDM, Inc.) was used to record the IMU data. The IMUs use wireless synchronization to ensure multiple units collect data with a precision of better than ±1 ms. Participants were also fitted with six reflective markers to collect simultaneous motion capture data. Markers were fixed to the forehead, the bilateral mandibular condyle of the head, the sternum, and the bilateral acromion process of the trunk. Motion capture data were collected using a 12-camera Motion Analysis system (120 Hz, Raptor-E, Motion Analysis Co., Santa Rosa, CA, USA) and processed using Cortex

v6.2.3 (Motion Analysis, Co.). Motion capture was synchronized with the IMU recording using an APDM synchronization box.

For each participant, a static trial was captured to define head and trunk segment position and orientation. This process allowed for each segment coordinate system to be rotated about the mediolateral axis, such that the anterior-posterior axis lies in the horizontal plane for the head and trunk in the static pose. A state space model and Kalman filter were used for sensor fusion between accelerometer, gyroscope, and magnetometer sensor data of the IMU [17]. Angular velocities were extracted from the head and trunk IMUs corresponding to rotations in the transverse plane (for L/R) and sagittal plane (for U/D). Angular displacement was calculated by integrating the angular velocity of the head in the intended direction (L/R or U/D).

Optical data were filtered using a dual-pass second order Butterworth filter (6 Hz cut-off) and up-sampled to match the sampling rate of the IMU data. The head and trunk segments (defined in Table 2) used a right-hand coordinate system. Flexion, abduction, and axial rotation were decomposed using Euler angles. Segment angles for the rotations of interest were calculated and differentiated to estimate rotational velocities.



For both IMU and optical datasets, time series data were segmented into individual head turns, allowing the calculation of ROM and peak rotational velocity. For the walking and tandem walking trials, only the straight walking segments were included. Portions of the trial when participants were turning at the ends of the walking path were removed. Turns were detected using a threshold turn angle greater than 45◦ and a peak turn velocity greater than 15 degrees per second [18].
