Laboratory for Movement Analysis, Children's University Hospital Basel (UKBB), Switzerland.
J Biomech. 2012 Aug 31;45(13):2306-9. doi: 10.1016/j.jbiomech.2012.06.012. Epub 2012 Jul 7.
Experimental data in human movement science commonly consist of repeated measurements under comparable conditions. One can face the question how to identify a single trial, a set of trials, or erroneous trials from the entire data set. This study presents and evaluates a Selection Method for a Representative Trial (SMaRT) based on the Principal Component Analysis. SMaRT was tested on 1841 data sets containing 11 joint angle curves of gait analysis. The automatically detected characteristic trials were compared with the choice of three independent experts. SMaRT required 1.4s to analyse 100 data sets consisting of 8±3 trials each. The robustness against outliers reached 98.8% (standard visual control). We conclude that SMaRT is a powerful tool to determine a representative, uncontaminated trial in movement analysis data sets with multiple parameters.
人体运动科学中的实验数据通常由在可比条件下进行的重复测量组成。人们可能会面临如何从整个数据集识别单个试验、一组试验或错误试验的问题。本研究提出并评估了一种基于主成分分析的代表性试验选择方法（SMaRT）。SMaRT 对包含 11 个步态分析关节角度曲线的 1841 个数据集进行了测试。自动检测到的特征试验与三位独立专家的选择进行了比较。SMaRT 分析由 8±3 个试验组成的 100 个数据集用时 1.4 秒。对离群值的稳健性达到 98.8%（标准视觉控制）。我们得出结论，SMaRT 是一种强大的工具，可用于确定具有多个参数的运动分析数据集中代表性、无污染的试验。
