摘要
复杂环境包括光照变化、噪声等多种因素,会导致人体轮廓特征变得模糊,掩盖真实的人体姿态特征,导致复杂环境人体运动姿态估计精度下降,为此提出了循环神经网络下的复杂环境人体运动姿态估计方法。采用三轴加速度传感器采集复杂环境人体运动信号,建立信号划分目标函数对人体运动信号展开划分处理。计算人体运动信号信号的离散系数与曲线积分,从而确定人体运动姿态特征,确保复杂环境中可有效捕获人体关节点的变化情况。将人体运动姿态特征输入建立循环神经网络中,从而得出人体运动姿态估计结果。实验结果表明,所提方法具有较高的信号划分精度和人体运动姿态估计精度,可以在实际中得到进一步应用。
Complex environments,including various factors such as illumination variations and noise,can cause human contour features to become blurred,obscuring the true human posture features and reducing the accuracy of human motion posture estimation in complex environments.To this end,a method for human motion posture estimation in complex environments using recurrent neural networks is proposed.A three-axis accelerometer is used to collect human motion signals in complex environments.A signal partitioning objective function is established to partition and process the human motion signals.The discrete coefficients and curve integrals of human motion signals are calculated to determine the characteristics of human motion posture,ensuring that changes in human joints can be effectively captured in complex environments.The human motion posture characteristics are input into a recurrent neural network to estimate human motion posture.Experimental results demonstrate that the proposed method has high signal segmentation accuracy and human motion posture estimation accuracy,andcan be further applied in practice.
作者
李晓兰
张红玲
LI Xiao-lan;ZHANG Hong-ling(Yan'an University,Yan'an Shaanxi 716000,China)
出处
《计算机仿真》
2026年第1期372-376,共5页
Computer Simulation
基金
陕西省科技厅重点研发一般项目(2025SF-YBXM-148)。
关键词
循环神经网络
信号划分
人体运动姿态
离散系数
曲线积分
Recurrent neural network
Signal segmentation
Human motion posture
Discrete coefficient
Curve integral
