人体运动的数据采集系统及识别设计

人体运动的数据采集系统可实时监测人体日常活动,为人体运动健康提供了更科学的辅助。利用Shimmer无线可穿戴传感器设备采集运动数据,建立了监测人体运动的8个部位传感器节点模型,并构建了10个具有代表性的动作进行识别。用中值滤波算法实现对原始信号的过滤处理,采用固定滑动窗口分割法对数据进行分割处理,结合特征提取技术来获取包含特性的特征向量,选择主成分分析(PCA)技术对特征提取后的特征向量进行降维处理。同时,提出了多节点运动识别模型,利用支持向量机(SVM)、k近邻(kNN)以及最近距离中心算法(NCC)三种不同的算法对单独节点和组合节点的情况下的传感器数据进行分类识别。实验结果表明,所提出的人体运动数据采集系统具有良好的效果,在使用相同识别算法的前提下,多节点组合识别能获得比单节点识别更好的效果。

Data Collection System and Recognition Design of Human Motion

The data acquisition system of human motion can monitor human daily activities in real time,and provide more scientific assistance for human sports health. Shimmer wireless wearable sensor equipment is used to collect motion data,and eight sensor node models for monitoring human motion are established,and ten representative actions are constructed for recognition. The median filtering algorithm is used to filter the original signal. Fixed sliding window segmentation method is used to segment the data. Feature extraction technology is combined to obtain feature vectors containing features. Principal component analysis (PCA) technology is selected to reduce the dimension of feature vectors after feature extraction. At the same time,a multi-node motion recognition model is proposed. Three different algorithms,support vector machine (SVM),k-nearest neighbor (kNN) and nearest distance center algorithm (NCC),are used to classify and recognize sensor data with individual and combined nodes. The experimental results show that the proposed human motion data acquisition system has a good effect. Under the premise of using the same recognition algorithm,multi-node combination recognition can achieve better results than single-node recognition.