Optimizing Cut-Off Frequencies and Filter Orders for Dynamic Local Reference Frames for Human Gait Analysis in Straight-Line and Turning Tasks

Adopting a dynamic local reference frame that consistently aligns with anatomical axes, especially during activities involving body turns, is crucial for the effective interpretation of human gait analysis. In this paper, we limit analysis to straight-line walking and turning tasks. However, dynamic local reference frames usually exhibit mediolateral oscillations, leading to misalignment with anatomical axes, especially in straight-line tasks. A potential solution is to decrease the mediolateral oscillations by a low-pass filter. In this study, the main objective is to determine optimal cut-off frequencies and orders of the low-pass filters for different dynamic local reference frames, including the pelvis-oriented, vCoM-oriented, and average angular velocity-oriented dynamic local reference frames. The optimal cut-off frequencies of different local reference frames were determined to be 0.4~0.6 Hz. Second-order filters had an improved filtering performance compared to first-order filters, while increasing the filter order (third- and fourth-order) did not further improve the filtering performance. A zero-phase second-order (biquad) Butterworth low-pass filter with a cut-off frequency of 0.5 Hz was found to be a good tradeoff between improving the alignment between the global and dynamic local reference frames in the straight-line walking (StrW) task, having a small impact on the alignment of dynamic local reference frames between stride cycles during the pre- and post-turning periods while preserving the turning features during the turning period of the walking with 180-degree (TurW) task, and considering the redundancy of increasing the filter orders. The use of this low-pass filter would allow for adapting a dynamic local reference frame during daily life tasks. It should be noted that our analysis was based on an optical motion capture system, it is still essential to assess the accuracy of defining local reference frames using inertial measurement unit data in future work.