Methods to improve the reliability of wearable sensor gait data.

生物力学步态分析是分析运动表现或评估病理运动模式的最普遍的研究方法之一。我们之前的发现奖和发现加速器补充奖通过开发新的统计方法和新的软件程序开发了新的方法来提高步态运动学数据的准确性和可重复性。在这项研究取得成功的基础上，并基于过去几年可穿戴传感器技术的重大进展，我们现在开始了将我们的实验室转移到现实世界并重新定义生物力学步态研究的独特研究挑战。 可穿戴传感器，如加速度计，陀螺仪和磁力计，是便携式和负担得起的，并迅速成为生物力学步态研究的常见替代品。然而，随着生物力学领域开始将可穿戴传感器用于研究目的，需要解决几个限制，必须建立基础研究。该研究计划的目标是确保可穿戴传感器数据的有效性，可靠性和可重复性基于新的统计方法。这里描述的研究计划的首要假设是，无法控制许多外在因素（例如温度，地形，倾斜度变化等）。在真实世界的数据收集过程中，这将显著降低可穿戴传感器数据的日常可靠性，并随后影响我们测量有效生物力学步态模式的能力。 这一主要假设将被评估，并介绍新的解决方案，通过专注于三个具体目标，涉及原始和创新的统计方法。目标1将专注于开发识别步态事件的新方法，目标2将专注于可能影响整体分类准确性的新分割和特征提取方法，目标3将建立可靠测量步态模式所需的数据会话数量。 这个新的研究项目处于融合数据科学和步态生物力学领域的最前沿，可以分析大量的生物力学数据，探索非结构化或复杂的数据集，并开发预测模型，从而产生新的见解。我们的科学方法将利用我们成熟的NSERC资助的步态分析和可穿戴传感器研究，我们希望开发新的方法，作为未来生物力学研究的基础。现在，当我们开始我们新授予的NSERC CREATE可穿戴培训和研究合作（We-TRAC）培训计划时，当前的发现资助研究计划有助于确保我们的HQP和NSE研究人员能够获得最好的工具。这些工具将最终改善我们在现实环境中使用可穿戴传感器进行步态生物力学研究的进展。

Biomechanical gait analysis is one of the most ubiquitous research methods for analysing sport performance or evaluating pathologic movement patterns. Our previous Discovery grant, and Discovery Accelerator Supplement award, developed novel methods to improve the accuracy and repeatability of gait kinematic data through the development of novel statistical methods and new software programs. Building on the success of this research, and based on significant advances in wearable sensor technology over the past few years, we now embark on the unique research challenge of moving our laboratory out into the real-world and redefining biomechanical gait research.     Wearable sensors, such as accelerometers, gyroscopes, and magnetometers, are portable and affordable and are quickly becoming a common alternative for biomechanics gait research. However, as the field of biomechanics begins to embrace the use of wearable sensors for research purposes, several limitations needs to be addressed and foundational research must be established. The objective of this research program is to ensure that wearable sensor data are valid, reliable and repeatable based on novel statistical methods. The overarching hypothesis of the research program described here is that the inability to control many extrinsic factors (e.g. temperature, terrain, changes in inclination, etc.) during real-world data collections will significantly reduce the day-to-day reliability of wearable sensor data and subsequently affect our ability to measure valid biomechanical gait patterns.     This main hypothesis will be evaluated, and novel solutions introduced, by focusing on three Specific Aims involving original and innovative statistical methods. Aim 1 will focus on developing new methods for identifying gait events, Aim 2 will focus on new segmentation and feature extraction methods that can influence overall classification accuracy, and Aim 3 will establish the number of data sessions necessary for reliable measurements of gait patterns.     This novel research program is at the forefront of merging the fields of data science and gait biomechanics to analyze large quantities of biomechanical data, explore unstructured or complex data sets, and develop prediction models that will produce new insights. Our scientific approach will capitalize on our well-established NSERC-funded gait analysis and wearable sensor research and we expect to develop novel methods that will serve as the foundation for future biomechanics research. Now as we begin our newly awarded NSERC CREATE Wearable Training and Research Collaboration (We-TRAC) training program, the current Discovery Grant research program serves to ensure that our HQP and NSE researchers have access to the best tools. These tools will ultimately improve our progress in using wearable sensors for gait biomechanics research in real-world settings.