【摘要】 针对医疗康复和人机交互领域中下肢外骨骼机器人对人体步态识别的需求，提出了一种基于足压与姿态信息融合的步态相位识别方法。以足底压力分布和足部运动姿态为研究对象，构建出一套可穿戴式足部运动数据采集系统，并收集了平地行走、坡路行走和上楼梯3种步态信息。采用卷积神经网络分类算法对上述3种步态进行相位识别，平地行走、坡路行走和上楼梯3种步态相位识别率分别达到97.0%、97.4%、97.6%。通过与支持向量机和反向传播神经网络的步态相位识别效果进行对比，验证了基于卷积神经网络的步态相位识别方法的精确性，为下肢外骨骼机器人在智能化人机协作中的应用提供了重要支持。更多还原

【Abstract】 To address the demand for human gait recognition in lower limb exoskeleton robots within the fields of medical rehabilitation and human-computer interaction, a method for gait phase recognition based on the fusion of foot pressure and posture information was proposed. By taking plantar pressure distribution and foot movement posture as research objects, a wearable foot movement data acquisition system was constructed. The system collected three types of gait information: walking on level ground, walking on a slope and walking up stairs. A convolutional neural network classification algorithm was utilized for phase recognition of these three gaits. The phase recognition rates for walking on level ground, walking on slopes, and walking up stairs reached 97.0%, 97.4% and 97.6%, respectively. The accuracy of this gait phase recognition method is verified through a comparison with the gait phase recognition effects of support vector machines and back propagation neural networks. This provides significant support for the application of lower limb exoskeleton robots in intelligent human-robot collaboration.更多还原