University of Copenhagen, Faculty of Health and Medical Sciences, Department of Large Animal Sciences, Hojbakkegaard Allé 5, 2630 Taastrup, Denmark.
The Royal Veterinary College, Department of Clinical Sciences and Services, Hawkshead Lane, North Mymms, AL9 7TA, United Kingdom.
J Biomech. 2014 Mar 21;47(5):1220-4. doi: 10.1016/j.jbiomech.2013.12.018. Epub 2014 Jan 22.
This study aimed to create an evidence base for detection of stance-phase timings from motion capture in horses. The objective was to compare the accuracy (bias) and precision (SD) for five published algorithms for the detection of hoof-on and hoof-off using force plates as the reference standard. Six horses were walked and trotted over eight force plates surrounded by a synchronised 12-camera infrared motion capture system. The five algorithms (A-E) were based on: (A) horizontal velocity of the hoof; (B) Fetlock angle and horizontal hoof velocity; (C) horizontal displacement of the hoof relative to the centre of mass; (D) horizontal velocity of the hoof relative to the Centre of Mass and; (E) vertical acceleration of the hoof. A total of 240 stance phases in walk and 240 stance phases in trot were included in the assessment. Method D provided the most accurate and precise results in walk for stance phase duration with a bias of 4.1% for front limbs and 4.8% for hind limbs. For trot we derived a combination of method A for hoof-on and method E for hoof-off resulting in a bias of -6.2% of stance in the front limbs and method B for the hind limbs with a bias of 3.8% of stance phase duration. We conclude that motion capture yields accurate and precise detection of gait events for horses walking and trotting over ground and the results emphasise a need for different algorithms for front limbs versus hind limbs in trot.
本研究旨在为基于运动捕捉的马站立相时相检测创建一个证据基础。目的是比较五种基于力板作为参考标准的检测马的蹄着地和离地的算法的准确性（偏差）和精密度（SD）。六匹马在被同步的 12 个摄像头红外运动捕捉系统包围的八个力板上进行慢步和跑步。这五种算法（A-E）基于：（A）蹄的水平速度；（B）球节角度和水平蹄速度；（C）蹄相对于质心的水平位移；（D）蹄相对于质心的水平速度；（E）蹄的垂直加速度。共评估了 240 个慢步和 240 个跑步的站立相。在慢步中，方法 D 提供了最准确和最精确的站立相持续时间结果，前肢的偏差为 4.1%，后肢的偏差为 4.8%。对于跑步，我们得出了一种方法 A 用于前脚着地，方法 E 用于后脚离地的组合，导致前脚的站立相偏差为-6.2%，后腿的方法 B 用于站立相持续时间，偏差为 3.8%。我们得出的结论是，运动捕捉可以准确和精确地检测马在地面上慢步和跑步的步态事件，结果强调了在跑步时需要针对前肢和后肢使用不同的算法。
