基于Mediapipe的下肢特征分析与步态周期检测研究  

Mediapipe-based lower limb characterization and gait cycle detection study

作　　者：罗子昂 吴钦木[1] LUO Zi′ang;WU Qinmu(College of Electrical Engineering,Guizhou University,Guiyang 550025,China)

机构地区：[1]贵州大学电气工程学院,贵阳550025

出　　处：《智能计算机与应用》2025年第3期133-139,共7页Intelligent Computer and Applications

基　　金：黔科合支撑[2021]一般442。

摘　　要：非接触式的步态周期检测无需人体与任何设备直接接触,只需通过传感器或摄像设备进行监测,减少了操作的复杂性和人体的不适感。为此提出了一种非接触式的步态周期检测方法。首先,以MediaPipe的Pose模型为基础对人体行走视频进行识别并提取出视频序列中的人体姿态拓扑图,提高了处理效率和实时性;然后,在姿势地标模型中分析不同的身体地标位置点组合成的特征所具有的周期性动态变化规律,将人体的两髋关节点与左右膝关节组成三角形面积比值和左腿膝关节角度值作为特征进行提取;最后,用三次样条插值法将视频各帧采集到的离散数据点拟合为具有周期性的连续曲线,并用切比雪夫滤波器对曲线进行滤波处理。通过实验验证,本文选取的下肢特征对于多视角步态周期检测具有良好的鲁棒性和准确率。Non-contact gait cycle detection does not require direct contact between the human body and any equipment,but only monitoring by sensors or camera equipment,which reduces the complexity of the operation and the discomfort of the human body.Regarding this situation,a non-contact gait cycle detection method is proposed.Based on the Pose model of MediaPipe,the human walking video is recognized and the topology of human posture in the video sequence is extracted,which improves the processing efficiency and real-time performance.Then,the periodic dynamic change law of the features combined with different body landmarks is analyzed in the pose landmark model,and the ratio of the area of the triangle composed of the two hip joints of human body to the left and right knee joints and the angle value of the left leg and knee joint are extracted as the features.Finally,the discrete data points collected from each frame of the video are fitted into a continuous curve with periodicity by the cubic spline interpolation method,and the curve is filtered by the Chebyshev filter.Through experimental verification,the lower limb features selected in this paper have good robustness and accuracy for multi-view gait cycle detection.

关 键 词：计算机应用 步态周期检测 特征选择 步态识别 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]