| 摘 要: 针对现有步态分析中步幅长度计算方法的现状,对基于足部惯性传感器的步幅估计方法进行了综述。首先,阐释了步态周期的定义,探讨了基于加速度计、陀螺仪的信号分割方法;其次,重点总结了传统的二重积分、基于经验模型以及人工智能模型的方法。研究结果表明,传统的二重积分方法精度高,适用于临床和运动训练等场景;基于经验模型的方法具备较强的实时处理能力,适用于行人航迹推算等对实时性要求较高的场景;基于人工智能模型的方法在处理复杂步态数据处理方面表现出色,但需大量数据和计算资源支持。未来研究应聚焦于数据融
合技术与实时处理算法的创新,以充分挖掘步幅测量技术在各领域的应用潜力。 |
| 关键词: 惯性传感器 步幅分割 步幅估计 二重积分 经验模型 机器学习 深度学习 |
|
中图分类号: TP391
文献标识码: A
|
|
| A Review of Stride Length Estimation Methods Based on Foot-Worn Inertial Sensor |
|
WAN Pengbo, ZHAO Yizhu, SHI Yujiao, LI Xueqing
|
(School of Art & Design, Shaanxi University of Science & Technology, Xi’an 710022, China)
Wanpengbo@yeah.net; 1394721662@qq.com; Syj13892681635@163.com; 2929696278@qq.com
|
| Abstract: This paper reviews stride length estimation methods based on foo-t worn inertial sensors in response to existing gait analysis techniques. Firstly, the definition of the gait cycle and signal segmentation approaches using accelerometers and gyroscopes are analyzed. Secondly, Key methods are summarized,including traditional double integration, empirical models, and artificial intelligence-based models. Results indicate that double integration offers high accuracy and is suitable for clinical and sports training scenarios; empirical models are ideal for rea-l time applications like pedestrian dead reckoning; while artifical intelligence-based models excel in handling complex gait data but require substantial computational resources and data support. Future research should focus on data fusion and rea-l time processing to enhance the practical applications of stride length measurement. |
| Keywords: inertial sensor stride segmentation stride length estimation double integration empirical model machine learning deep learning |
