基于天顶轴速度零点的行人步数检测

Pedestrians’Gait Detection Based on Zenith Axis Velocity Zeros

针对现有峰值检测算法适用性较低和深度学习方法算力要求较高的问题,研究和完整阐述了步态规律和数据特征,通过分析步态规律提出了基于天顶轴速度零值的行人步数检测算法。通过将载体坐标系下的加速度计数据转换到地理坐标系下,使用经验模态分解、趋势项剔除等方法提取出每一步中的有效天顶轴速度零点,统计有效零点的个数作为步数。在多种行走状态下的实验表明,相比改进的峰值检测模型,本模型的准确率与可靠性有明显提升,在行人过快或过慢行走时提升效果更为显著。

To address the problems of low applicability of ex⁃isting peak detection algorithms and high arithmetic require⁃ments of deep learning methods,the gait law and data charac⁃teristics are studied and completely elaborated,and a pedestri⁃ans’gait detection algorithm based on zenith axis velocity zeros is proposed by the gait law.The effective zenith axis velocity zeros in each gait are extracted by converting the ac⁃celerometer data under the body frame to the geographic frame,performing empirical modal decomposition,trend item rejection in turn,finally the number of effective zeros is count⁃ed as the number of gaits.Experiments under multiple walk⁃ing states show that the accuracy and reliability of the present model are significantly improved compared to the improved peak detection model,and the improvement can be more significant when pedestrians are walking too fast or too slow.